Face recognition improvements (#16034)

* Start working on bird processor

* Initial setup for bird processing

* Improvements to handling

* Get classification working

* Cleanup classification

* Add classification config

* Update sort

.cspell/frigate-dictionary.txt

@@ -2,6 +2,7 @@ aarch
 absdiff
 airockchip
 Alloc
+alpr
 Amcrest
 amdgpu
 analyzeduration
@@ -61,6 +62,7 @@ dsize
 dtype
 ECONNRESET
 edgetpu
+facenet
 fastapi
 faststart
 fflags
@@ -114,6 +116,8 @@ itemsize
 Jellyfin
 jetson
 jetsons
+jina
+jinaai
 joserfc
 jsmpeg
 jsonify
@@ -187,6 +191,7 @@ openai
 opencv
 openvino
 OWASP
+paddleocr
 paho
 passwordless
 popleft
@@ -308,4 +313,4 @@ yolo
 yolonas
 yolox
 zeep
-zerolatency
+zerolatency

.github/workflows/ci.yml

@@ -75,15 +75,6 @@ jobs:
             rpi.tags=${{ steps.setup.outputs.image-name }}-rpi
             *.cache-from=type=registry,ref=${{ steps.setup.outputs.cache-name }}-arm64
             *.cache-to=type=registry,ref=${{ steps.setup.outputs.cache-name }}-arm64,mode=max
-      - name: Build and push Rockchip build
-        uses: docker/bake-action@v3
-        with:
-          push: true
-          targets: rk
-          files: docker/rockchip/rk.hcl
-          set: |
-            rk.tags=${{ steps.setup.outputs.image-name }}-rk
-            *.cache-from=type=gha
   jetson_jp4_build:
     runs-on: ubuntu-latest
     name: Jetson Jetpack 4

.github/workflows/pull_request.yml

@@ -6,7 +6,7 @@ on:
       - "docs/**"
 
 env:
-  DEFAULT_PYTHON: 3.9
+  DEFAULT_PYTHON: 3.11
 
 jobs:
   build_devcontainer:
@@ -76,7 +76,7 @@ jobs:
         with:
           persist-credentials: false
       - name: Set up Python ${{ env.DEFAULT_PYTHON }}
-        uses: actions/setup-python@v5.1.0
+        uses: actions/setup-python@v5.3.0
         with:
           python-version: ${{ env.DEFAULT_PYTHON }}
       - name: Install requirements

Makefile

@@ -1,7 +1,7 @@
 default_target: local
 
 COMMIT_HASH := $(shell git log -1 --pretty=format:"%h"|tail -1)
-VERSION = 0.15.0
+VERSION = 0.16.0
 IMAGE_REPO ?= ghcr.io/blakeblackshear/frigate
 GITHUB_REF_NAME ?= $(shell git rev-parse --abbrev-ref HEAD)
 BOARDS= #Initialized empty

benchmark.py

@@ -61,7 +61,7 @@ def start(id, num_detections, detection_queue, event):
     object_detector.cleanup()
     print(f"{id} - Processed for {duration:.2f} seconds.")
     print(f"{id} - FPS: {object_detector.fps.eps():.2f}")
-    print(f"{id} - Average frame processing time: {mean(frame_times)*1000:.2f}ms")
+    print(f"{id} - Average frame processing time: {mean(frame_times) * 1000:.2f}ms")
 
 
 ######

docker/hailo8l/Dockerfile

@@ -5,6 +5,7 @@ ARG DEBIAN_FRONTEND=noninteractive
 # Build Python wheels
 FROM wheels AS h8l-wheels
 
+RUN python3 -m pip config set global.break-system-packages true
 COPY docker/main/requirements-wheels.txt /requirements-wheels.txt
 COPY docker/hailo8l/requirements-wheels-h8l.txt /requirements-wheels-h8l.txt
 
@@ -30,6 +31,7 @@ COPY --from=hailort /hailo-wheels /deps/hailo-wheels
 COPY --from=hailort /rootfs/ /
 
 # Install the wheels
+RUN python3 -m pip config set global.break-system-packages true
 RUN pip3 install -U /deps/h8l-wheels/*.whl
 RUN pip3 install -U /deps/hailo-wheels/*.whl
 

docker/hailo8l/install_hailort.sh

@@ -2,7 +2,7 @@
 
 set -euxo pipefail
 
-hailo_version="4.19.0"
+hailo_version="4.20.0"
 
 if [[ "${TARGETARCH}" == "amd64" ]]; then
     arch="x86_64"
@@ -15,5 +15,5 @@ wget -qO- "https://github.com/frigate-nvr/hailort/releases/download/v${hailo_ver
 
 mkdir -p /hailo-wheels
 
-wget -P /hailo-wheels/ "https://github.com/frigate-nvr/hailort/releases/download/v${hailo_version}/hailort-${hailo_version}-cp39-cp39-linux_${arch}.whl"
+wget -P /hailo-wheels/ "https://github.com/frigate-nvr/hailort/releases/download/v${hailo_version}/hailort-${hailo_version}-cp311-cp311-linux_${arch}.whl"
 

docker/hailo8l/requirements-wheels-h8l.txt

@@ -1,12 +1,12 @@
-appdirs==1.4.4
-argcomplete==2.0.0
-contextlib2==0.6.0.post1
-distlib==0.3.6
-filelock==3.8.0
-future==0.18.2
-importlib-metadata==5.1.0
-importlib-resources==5.1.2
-netaddr==0.8.0
-netifaces==0.10.9
-verboselogs==1.7
-virtualenv==20.17.0
+appdirs==1.4.*
+argcomplete==2.0.*
+contextlib2==0.6.*
+distlib==0.3.*
+filelock==3.8.*
+future==0.18.*
+importlib-metadata==5.1.*
+importlib-resources==5.1.*
+netaddr==0.8.*
+netifaces==0.10.*
+verboselogs==1.7.*
+virtualenv==20.17.*

docker/hailo8l/user_installation.sh

@@ -4,6 +4,7 @@
 sudo apt-get update
 sudo apt-get install -y build-essential cmake git wget
 
+hailo_version="4.20.0"
 arch=$(uname -m)
 
 if [[ $arch == "x86_64" ]]; then
@@ -13,7 +14,7 @@ else
 fi
 
 # Clone the HailoRT driver repository
-git clone --depth 1 --branch v4.19.0 https://github.com/hailo-ai/hailort-drivers.git
+git clone --depth 1 --branch v${hailo_version} https://github.com/hailo-ai/hailort-drivers.git
 
 # Build and install the HailoRT driver
 cd hailort-drivers/linux/pcie

docker/main/Dockerfile

@@ -3,12 +3,12 @@
 # https://askubuntu.com/questions/972516/debian-frontend-environment-variable
 ARG DEBIAN_FRONTEND=noninteractive
 
-ARG BASE_IMAGE=debian:11
-ARG SLIM_BASE=debian:11-slim
+ARG BASE_IMAGE=debian:12
+ARG SLIM_BASE=debian:12-slim
 
 FROM ${BASE_IMAGE} AS base
 
-FROM --platform=${BUILDPLATFORM} debian:11 AS base_host
+FROM --platform=${BUILDPLATFORM} debian:12 AS base_host
 
 FROM ${SLIM_BASE} AS slim-base
 
@@ -66,8 +66,8 @@ COPY docker/main/requirements-ov.txt /requirements-ov.txt
 RUN apt-get -qq update \
     && apt-get -qq install -y wget python3 python3-dev python3-distutils gcc pkg-config libhdf5-dev \
     && wget -q https://bootstrap.pypa.io/get-pip.py -O get-pip.py \
-    && python3 get-pip.py "pip" \
-    && pip install -r /requirements-ov.txt
+    && python3 get-pip.py "pip" --break-system-packages \
+    && pip install --break-system-packages -r /requirements-ov.txt
 
 # Get OpenVino Model
 RUN --mount=type=bind,source=docker/main/build_ov_model.py,target=/build_ov_model.py \
@@ -139,24 +139,17 @@ ARG TARGETARCH
 # Use a separate container to build wheels to prevent build dependencies in final image
 RUN apt-get -qq update \
     && apt-get -qq install -y \
-    apt-transport-https \
-    gnupg \
-    wget \
-    # the key fingerprint can be obtained from https://ftp-master.debian.org/keys.html
-    && wget -qO- "https://keyserver.ubuntu.com/pks/lookup?op=get&search=0xA4285295FC7B1A81600062A9605C66F00D6C9793" | \
-    gpg --dearmor > /usr/share/keyrings/debian-archive-bullseye-stable.gpg \
-    && echo "deb [signed-by=/usr/share/keyrings/debian-archive-bullseye-stable.gpg] http://deb.debian.org/debian bullseye main contrib non-free" | \
-    tee /etc/apt/sources.list.d/debian-bullseye-nonfree.list \
+    apt-transport-https wget \
     && apt-get -qq update \
     && apt-get -qq install -y \
-    python3.9 \
-    python3.9-dev \
+    python3 \
+    python3-dev \
     # opencv dependencies
     build-essential cmake git pkg-config libgtk-3-dev \
     libavcodec-dev libavformat-dev libswscale-dev libv4l-dev \
     libxvidcore-dev libx264-dev libjpeg-dev libpng-dev libtiff-dev \
     gfortran openexr libatlas-base-dev libssl-dev\
-    libtbb2 libtbb-dev libdc1394-22-dev libopenexr-dev \
+    libtbbmalloc2 libtbb-dev libdc1394-dev libopenexr-dev \
     libgstreamer-plugins-base1.0-dev libgstreamer1.0-dev \
     # sqlite3 dependencies
     tclsh \
@@ -164,14 +157,11 @@ RUN apt-get -qq update \
     gcc gfortran libopenblas-dev liblapack-dev && \
     rm -rf /var/lib/apt/lists/*
 
-# Ensure python3 defaults to python3.9
-RUN update-alternatives --install /usr/bin/python3 python3 /usr/bin/python3.9 1
-
 RUN wget -q https://bootstrap.pypa.io/get-pip.py -O get-pip.py \
-    && python3 get-pip.py "pip"
+    && python3 get-pip.py "pip" --break-system-packages
 
 COPY docker/main/requirements.txt /requirements.txt
-RUN pip3 install -r /requirements.txt
+RUN pip3 install -r /requirements.txt --break-system-packages
 
 # Build pysqlite3 from source
 COPY docker/main/build_pysqlite3.sh /build_pysqlite3.sh
@@ -222,8 +212,8 @@ RUN --mount=type=bind,source=docker/main/install_deps.sh,target=/deps/install_de
     /deps/install_deps.sh
 
 RUN --mount=type=bind,from=wheels,source=/wheels,target=/deps/wheels \
-    python3 -m pip install --upgrade pip && \
-    pip3 install -U /deps/wheels/*.whl
+    python3 -m pip install --upgrade pip --break-system-packages && \
+    pip3 install -U /deps/wheels/*.whl --break-system-packages
 
 COPY --from=deps-rootfs / /
 
@@ -270,7 +260,7 @@ RUN apt-get update \
     && rm -rf /var/lib/apt/lists/*
 
 RUN --mount=type=bind,source=./docker/main/requirements-dev.txt,target=/workspace/frigate/requirements-dev.txt \
-    pip3 install -r requirements-dev.txt
+    pip3 install -r requirements-dev.txt --break-system-packages
 
 HEALTHCHECK NONE
 

docker/main/build_nginx.sh

@@ -8,8 +8,7 @@ SECURE_TOKEN_MODULE_VERSION="1.5"
 SET_MISC_MODULE_VERSION="v0.33"
 NGX_DEVEL_KIT_VERSION="v0.3.3"
 
-cp /etc/apt/sources.list /etc/apt/sources.list.d/sources-src.list
-sed -i 's|deb http|deb-src http|g' /etc/apt/sources.list.d/sources-src.list
+sed -i '/^Types:/s/deb/& deb-src/' /etc/apt/sources.list.d/debian.sources
 apt-get update
 
 apt-get -yqq build-dep nginx

docker/main/build_ov_model.py

@@ -4,7 +4,7 @@ from openvino.tools import mo
 ov_model = mo.convert_model(
     "/models/ssdlite_mobilenet_v2_coco_2018_05_09/frozen_inference_graph.pb",
     compress_to_fp16=True,
-    transformations_config="/usr/local/lib/python3.9/dist-packages/openvino/tools/mo/front/tf/ssd_v2_support.json",
+    transformations_config="/usr/local/lib/python3.11/dist-packages/openvino/tools/mo/front/tf/ssd_v2_support.json",
     tensorflow_object_detection_api_pipeline_config="/models/ssdlite_mobilenet_v2_coco_2018_05_09/pipeline.config",
     reverse_input_channels=True,
 )

docker/main/build_sqlite_vec.sh

@@ -4,8 +4,7 @@ set -euxo pipefail
 
 SQLITE_VEC_VERSION="0.1.3"
 
-cp /etc/apt/sources.list /etc/apt/sources.list.d/sources-src.list
-sed -i 's|deb http|deb-src http|g' /etc/apt/sources.list.d/sources-src.list
+sed -i '/^Types:/s/deb/& deb-src/' /etc/apt/sources.list.d/debian.sources
 apt-get update
 apt-get -yqq build-dep sqlite3 gettext git
 

docker/main/install_deps.sh

@@ -11,33 +11,34 @@ apt-get -qq install --no-install-recommends -y \
     lbzip2 \
     procps vainfo \
     unzip locales tzdata libxml2 xz-utils \
-    python3.9 \
+    python3 \
     python3-pip \
     curl \
     lsof \
     jq \
-    nethogs
-
-# ensure python3 defaults to python3.9
-update-alternatives --install /usr/bin/python3 python3 /usr/bin/python3.9 1
+    nethogs \
+    libgl1 \
+    libglib2.0-0 \
+    libusb-1.0.0
 
 mkdir -p -m 600 /root/.gnupg
 
-# add coral repo
-curl -fsSLo - https://packages.cloud.google.com/apt/doc/apt-key.gpg | \
-    gpg --dearmor -o /etc/apt/trusted.gpg.d/google-cloud-packages-archive-keyring.gpg
-echo "deb https://packages.cloud.google.com/apt coral-edgetpu-stable main" | tee /etc/apt/sources.list.d/coral-edgetpu.list
-echo "libedgetpu1-max libedgetpu/accepted-eula select true" | debconf-set-selections
+# install coral runtime
+wget -q -O /tmp/libedgetpu1-max.deb "https://github.com/feranick/libedgetpu/releases/download/16.0TF2.17.0-1/libedgetpu1-max_16.0tf2.17.0-1.bookworm_${TARGETARCH}.deb"
+unset DEBIAN_FRONTEND
+yes | dpkg -i /tmp/libedgetpu1-max.deb && export DEBIAN_FRONTEND=noninteractive
+rm /tmp/libedgetpu1-max.deb
 
-# enable non-free repo in Debian
-if grep -q "Debian" /etc/issue; then
-    sed -i -e's/ main/ main contrib non-free/g' /etc/apt/sources.list
+# install python3 & tflite runtime
+if [[ "${TARGETARCH}" == "amd64" ]]; then
+    pip3 install --break-system-packages https://github.com/feranick/TFlite-builds/releases/download/v2.17.0/tflite_runtime-2.17.0-cp311-cp311-linux_x86_64.whl
+    pip3 install --break-system-packages https://github.com/feranick/pycoral/releases/download/2.0.2TF2.17.0/pycoral-2.0.2-cp311-cp311-linux_x86_64.whl
 fi
 
-# coral drivers
-apt-get -qq update
-apt-get -qq install --no-install-recommends --no-install-suggests -y \
-    libedgetpu1-max python3-tflite-runtime python3-pycoral
+if [[ "${TARGETARCH}" == "arm64" ]]; then
+    pip3 install --break-system-packages https://github.com/feranick/TFlite-builds/releases/download/v2.17.0/tflite_runtime-2.17.0-cp311-cp311-linux_aarch64.whl
+    pip3 install --break-system-packages https://github.com/feranick/pycoral/releases/download/2.0.2TF2.17.0/pycoral-2.0.2-cp311-cp311-linux_aarch64.whl
+fi
 
 # btbn-ffmpeg -> amd64
 if [[ "${TARGETARCH}" == "amd64" ]]; then
@@ -65,23 +66,15 @@ fi
 
 # arch specific packages
 if [[ "${TARGETARCH}" == "amd64" ]]; then
-    # use debian bookworm for amd / intel-i965 driver packages
-    echo 'deb https://deb.debian.org/debian bookworm main contrib non-free' >/etc/apt/sources.list.d/debian-bookworm.list
-    apt-get -qq update
+    # install amd / intel-i965 driver packages
     apt-get -qq install --no-install-recommends --no-install-suggests -y \
         i965-va-driver intel-gpu-tools onevpl-tools \
         libva-drm2 \
         mesa-va-drivers radeontop
 
-    # something about this dependency requires it to be installed in a separate call rather than in the line above
-    apt-get -qq install --no-install-recommends --no-install-suggests -y \
-        i965-va-driver-shaders
-
     # intel packages use zst compression so we need to update dpkg
     apt-get install -y dpkg
 
-    rm -f /etc/apt/sources.list.d/debian-bookworm.list
-
     # use intel apt intel packages
     wget -qO - https://repositories.intel.com/gpu/intel-graphics.key | gpg --yes --dearmor --output /usr/share/keyrings/intel-graphics.gpg
     echo "deb [arch=amd64 signed-by=/usr/share/keyrings/intel-graphics.gpg] https://repositories.intel.com/gpu/ubuntu jammy client" | tee /etc/apt/sources.list.d/intel-gpu-jammy.list

docker/main/requirements-wheels.txt

@@ -10,10 +10,10 @@ imutils == 0.5.*
 joserfc == 1.0.*
 pathvalidate == 3.2.*
 markupsafe == 2.1.*
+python-multipart == 0.0.12
+# General
 mypy == 1.6.1
-numpy == 1.26.*
-onvif_zeep == 0.2.12
-opencv-python-headless == 4.9.0.*
+onvif-zeep-async == 3.1.*
 paho-mqtt == 2.1.*
 pandas == 2.2.*
 peewee == 3.17.*
@@ -27,15 +27,19 @@ ruamel.yaml == 0.18.*
 tzlocal == 5.2
 requests == 2.32.*
 types-requests == 2.32.*
-scipy == 1.13.*
 norfair == 2.2.*
 setproctitle == 1.3.*
 ws4py == 0.5.*
 unidecode == 1.3.*
+# Image Manipulation
+numpy == 1.26.*
+opencv-python-headless == 4.10.0.*
+opencv-contrib-python == 4.9.0.*
+scipy == 1.14.*
 # OpenVino & ONNX
-openvino == 2024.3.*
-onnxruntime-openvino == 1.19.* ; platform_machine == 'x86_64'
-onnxruntime == 1.19.* ; platform_machine == 'aarch64'
+openvino == 2024.4.*
+onnxruntime-openvino == 1.20.* ; platform_machine == 'x86_64'
+onnxruntime == 1.20.* ; platform_machine == 'aarch64'
 # Embeddings
 transformers == 4.45.*
 # Generative AI
@@ -45,3 +49,6 @@ openai == 1.51.*
 # push notifications
 py-vapid == 1.9.*
 pywebpush == 2.0.*
+# alpr
+pyclipper == 1.3.*
+shapely == 2.0.*

docker/main/requirements.txt

@@ -1,2 +1,2 @@
-scikit-build == 0.17.*
+scikit-build == 0.18.*
 nvidia-pyindex

docker/main/rootfs/usr/local/nginx/conf/nginx.conf

@@ -81,6 +81,9 @@ http {
         open_file_cache_errors on;
         aio on;
 
+        # file upload size
+        client_max_body_size 10M;
+
         # https://github.com/kaltura/nginx-vod-module#vod_open_file_thread_pool
         vod_open_file_thread_pool default;
 

docker/rockchip/COCO/coco_subset_20.txt

@@ -0,0 +1,20 @@
+./subset/000000005001.jpg
+./subset/000000038829.jpg
+./subset/000000052891.jpg
+./subset/000000075612.jpg
+./subset/000000098261.jpg
+./subset/000000181542.jpg
+./subset/000000215245.jpg
+./subset/000000277005.jpg
+./subset/000000288685.jpg
+./subset/000000301421.jpg
+./subset/000000334371.jpg
+./subset/000000348481.jpg
+./subset/000000373353.jpg
+./subset/000000397681.jpg
+./subset/000000414673.jpg
+./subset/000000419312.jpg
+./subset/000000465822.jpg
+./subset/000000475732.jpg
+./subset/000000559707.jpg
+./subset/000000574315.jpg

docker/rockchip/Dockerfile

@@ -7,21 +7,26 @@ FROM wheels as rk-wheels
 COPY docker/main/requirements-wheels.txt /requirements-wheels.txt
 COPY docker/rockchip/requirements-wheels-rk.txt /requirements-wheels-rk.txt
 RUN sed -i "/https:\/\//d" /requirements-wheels.txt
+RUN sed -i "/onnxruntime/d" /requirements-wheels.txt
+RUN python3 -m pip config set global.break-system-packages true
 RUN pip3 wheel --wheel-dir=/rk-wheels -c /requirements-wheels.txt -r /requirements-wheels-rk.txt
+RUN rm -rf /rk-wheels/opencv_python-*
 
 FROM deps AS rk-frigate
 ARG TARGETARCH
 
 RUN --mount=type=bind,from=rk-wheels,source=/rk-wheels,target=/deps/rk-wheels \
-    pip3 install -U /deps/rk-wheels/*.whl
+    pip3 install --no-deps -U /deps/rk-wheels/*.whl --break-system-packages
 
 WORKDIR /opt/frigate/
 COPY --from=rootfs / /
+COPY docker/rockchip/COCO /COCO
+COPY docker/rockchip/conv2rknn.py /opt/conv2rknn.py
 
-ADD https://github.com/MarcA711/rknn-toolkit2/releases/download/v2.0.0/librknnrt.so /usr/lib/
+ADD https://github.com/MarcA711/rknn-toolkit2/releases/download/v2.3.0/librknnrt.so /usr/lib/
 
 RUN rm -rf /usr/lib/btbn-ffmpeg/bin/ffmpeg
 RUN rm -rf /usr/lib/btbn-ffmpeg/bin/ffprobe
-ADD --chmod=111 https://github.com/MarcA711/Rockchip-FFmpeg-Builds/releases/download/6.1-5/ffmpeg /usr/lib/ffmpeg/6.0/bin/
-ADD --chmod=111 https://github.com/MarcA711/Rockchip-FFmpeg-Builds/releases/download/6.1-5/ffprobe /usr/lib/ffmpeg/6.0/bin/
+ADD --chmod=111 https://github.com/MarcA711/Rockchip-FFmpeg-Builds/releases/download/6.1-6/ffmpeg /usr/lib/ffmpeg/6.0/bin/
+ADD --chmod=111 https://github.com/MarcA711/Rockchip-FFmpeg-Builds/releases/download/6.1-6/ffprobe /usr/lib/ffmpeg/6.0/bin/
 ENV PATH="/usr/lib/ffmpeg/6.0/bin/:${PATH}"

docker/rockchip/conv2rknn.py

@@ -0,0 +1,82 @@
+import os
+
+import rknn
+import yaml
+from rknn.api import RKNN
+
+try:
+    with open(rknn.__path__[0] + "/VERSION") as file:
+        tk_version = file.read().strip()
+except FileNotFoundError:
+    pass
+
+try:
+    with open("/config/conv2rknn.yaml", "r") as config_file:
+        configuration = yaml.safe_load(config_file)
+except FileNotFoundError:
+    raise Exception("Please place a config.yaml file in /config/conv2rknn.yaml")
+
+if configuration["config"] != None:
+    rknn_config = configuration["config"]
+else:
+    rknn_config = {}
+
+if not os.path.isdir("/config/model_cache/rknn_cache/onnx"):
+    raise Exception(
+        "Place the onnx models you want to convert to rknn format in /config/model_cache/rknn_cache/onnx"
+    )
+
+if "soc" not in configuration:
+    try:
+        with open("/proc/device-tree/compatible") as file:
+            soc = file.read().split(",")[-1].strip("\x00")
+    except FileNotFoundError:
+        raise Exception("Make sure to run docker in privileged mode.")
+
+    configuration["soc"] = [
+        soc,
+    ]
+
+if "quantization" not in configuration:
+    configuration["quantization"] = False
+
+if "output_name" not in configuration:
+    configuration["output_name"] = "{{input_basename}}"
+
+for input_filename in os.listdir("/config/model_cache/rknn_cache/onnx"):
+    for soc in configuration["soc"]:
+        quant = "i8" if configuration["quantization"] else "fp16"
+
+        input_path = "/config/model_cache/rknn_cache/onnx/" + input_filename
+        input_basename = input_filename[: input_filename.rfind(".")]
+
+        output_filename = (
+            configuration["output_name"].format(
+                quant=quant,
+                input_basename=input_basename,
+                soc=soc,
+                tk_version=tk_version,
+            )
+            + ".rknn"
+        )
+        output_path = "/config/model_cache/rknn_cache/" + output_filename
+
+        rknn_config["target_platform"] = soc
+
+        rknn = RKNN(verbose=True)
+        rknn.config(**rknn_config)
+
+        if rknn.load_onnx(model=input_path) != 0:
+            raise Exception("Error loading model.")
+
+        if (
+            rknn.build(
+                do_quantization=configuration["quantization"],
+                dataset="/COCO/coco_subset_20.txt",
+            )
+            != 0
+        ):
+            raise Exception("Error building model.")
+
+        if rknn.export_rknn(output_path) != 0:
+            raise Exception("Error exporting rknn model.")

docker/rockchip/requirements-wheels-rk.txt

@@ -1 +1,2 @@
-rknn-toolkit-lite2 @ https://github.com/MarcA711/rknn-toolkit2/releases/download/v2.0.0/rknn_toolkit_lite2-2.0.0b0-cp39-cp39-linux_aarch64.whl
+rknn-toolkit2 == 2.3.0
+rknn-toolkit-lite2 == 2.3.0

docker/rocm/Dockerfile

@@ -34,7 +34,7 @@ RUN mkdir -p /opt/rocm-dist/etc/ld.so.conf.d/
 RUN echo /opt/rocm/lib|tee /opt/rocm-dist/etc/ld.so.conf.d/rocm.conf
 
 #######################################################################
-FROM --platform=linux/amd64 debian:11 as debian-base
+FROM --platform=linux/amd64 debian:12 as debian-base
 
 RUN apt-get update && apt-get -y upgrade
 RUN apt-get -y install --no-install-recommends libelf1 libdrm2 libdrm-amdgpu1 libnuma1 kmod
@@ -51,7 +51,7 @@ COPY --from=rocm /opt/rocm-$ROCM /opt/rocm-$ROCM
 RUN ln -s /opt/rocm-$ROCM /opt/rocm
 
 RUN apt-get -y install g++ cmake
-RUN apt-get -y install python3-pybind11 python3.9-distutils python3-dev
+RUN apt-get -y install python3-pybind11 python3-distutils python3-dev
 
 WORKDIR /opt/build
 
@@ -70,10 +70,11 @@ RUN apt-get -y install libnuma1
 WORKDIR /opt/frigate/
 COPY --from=rootfs / /
 
-COPY docker/rocm/requirements-wheels-rocm.txt /requirements.txt
-RUN python3 -m pip install --upgrade pip \
-    && pip3 uninstall -y onnxruntime-openvino \
-    && pip3 install -r /requirements.txt
+# Temporarily disabled to see if a new wheel can be built to support py3.11
+#COPY docker/rocm/requirements-wheels-rocm.txt /requirements.txt
+#RUN python3 -m pip install --upgrade pip \
+#    && pip3 uninstall -y onnxruntime-openvino \
+#    && pip3 install -r /requirements.txt
 
 #######################################################################
 FROM scratch AS rocm-dist
@@ -86,12 +87,12 @@ COPY --from=rocm /opt/rocm-$ROCM/share/miopen/db/*$AMDGPU* /opt/rocm-$ROCM/share
 COPY --from=rocm /opt/rocm-$ROCM/share/miopen/db/*gfx908* /opt/rocm-$ROCM/share/miopen/db/
 COPY --from=rocm /opt/rocm-$ROCM/lib/rocblas/library/*$AMDGPU* /opt/rocm-$ROCM/lib/rocblas/library/
 COPY --from=rocm /opt/rocm-dist/ /
-COPY --from=debian-build /opt/rocm/lib/migraphx.cpython-39-x86_64-linux-gnu.so /opt/rocm-$ROCM/lib/
+COPY --from=debian-build /opt/rocm/lib/migraphx.cpython-311-x86_64-linux-gnu.so /opt/rocm-$ROCM/lib/
 
 #######################################################################
 FROM deps-prelim AS rocm-prelim-hsa-override0
-
-ENV HSA_ENABLE_SDMA=0
+\
+    ENV HSA_ENABLE_SDMA=0
 
 COPY --from=rocm-dist / /
 

docker/rpi/install_deps.sh

@@ -24,7 +24,7 @@ sed -i -e's/ main/ main contrib non-free/g' /etc/apt/sources.list
 if [[ "${TARGETARCH}" == "arm64" ]]; then
     # add raspberry pi repo
     gpg --no-default-keyring --keyring /usr/share/keyrings/raspbian.gpg --keyserver keyserver.ubuntu.com --recv-keys 82B129927FA3303E
-    echo "deb [signed-by=/usr/share/keyrings/raspbian.gpg] https://archive.raspberrypi.org/debian/ bullseye main" | tee /etc/apt/sources.list.d/raspi.list
+    echo "deb [signed-by=/usr/share/keyrings/raspbian.gpg] https://archive.raspberrypi.org/debian/ bookworm main" | tee /etc/apt/sources.list.d/raspi.list
     apt-get -qq update
     apt-get -qq install --no-install-recommends --no-install-suggests -y ffmpeg
 fi

docker/tensorrt/Dockerfile.amd64

@@ -7,33 +7,19 @@ ARG DEBIAN_FRONTEND=noninteractive
 FROM wheels as trt-wheels
 ARG DEBIAN_FRONTEND
 ARG TARGETARCH
+RUN python3 -m pip config set global.break-system-packages true
 
 # Add TensorRT wheels to another folder
 COPY docker/tensorrt/requirements-amd64.txt /requirements-tensorrt.txt
 RUN mkdir -p /trt-wheels && pip3 wheel --wheel-dir=/trt-wheels -r /requirements-tensorrt.txt
 
-# Build CuDNN
-FROM wget AS cudnn-deps
-
-ARG COMPUTE_LEVEL
-
-RUN apt-get update \
-    && apt-get install -y git build-essential
-
-RUN wget https://developer.download.nvidia.com/compute/cuda/repos/debian11/x86_64/cuda-keyring_1.1-1_all.deb \
-&& dpkg -i cuda-keyring_1.1-1_all.deb \
-&& apt-get update \
-&& apt-get -y install cuda-toolkit \
-&& rm -rf /var/lib/apt/lists/*
-
 FROM tensorrt-base AS frigate-tensorrt
-ENV TRT_VER=8.5.3
+ENV TRT_VER=8.6.1
+RUN python3 -m pip config set global.break-system-packages true
 RUN --mount=type=bind,from=trt-wheels,source=/trt-wheels,target=/deps/trt-wheels \
-    pip3 install -U /deps/trt-wheels/*.whl && \
+    pip3 install -U /deps/trt-wheels/*.whl --break-system-packages && \
     ldconfig
-COPY --from=cudnn-deps /usr/local/cuda-12.6 /usr/local/cuda
 
-ENV LD_LIBRARY_PATH=/usr/local/lib/python3.9/dist-packages/tensorrt:/usr/local/cuda/lib64:/usr/local/lib/python3.9/dist-packages/nvidia/cufft/lib
 WORKDIR /opt/frigate/
 COPY --from=rootfs / /
 
@@ -42,8 +28,8 @@ FROM devcontainer AS devcontainer-trt
 
 COPY --from=trt-deps /usr/local/lib/libyolo_layer.so /usr/local/lib/libyolo_layer.so
 COPY --from=trt-deps /usr/local/src/tensorrt_demos /usr/local/src/tensorrt_demos
-COPY --from=cudnn-deps /usr/local/cuda-12.6 /usr/local/cuda
+COPY --from=trt-deps /usr/local/cuda-12.1 /usr/local/cuda
 COPY docker/tensorrt/detector/rootfs/ /
 COPY --from=trt-deps /usr/local/lib/libyolo_layer.so /usr/local/lib/libyolo_layer.so
 RUN --mount=type=bind,from=trt-wheels,source=/trt-wheels,target=/deps/trt-wheels \
-    pip3 install -U /deps/trt-wheels/*.whl
+    pip3 install -U /deps/trt-wheels/*.whl --break-system-packages

docker/tensorrt/Dockerfile.arm64

@@ -41,11 +41,11 @@ RUN --mount=type=bind,source=docker/tensorrt/detector/build_python_tensorrt.sh,t
     && TENSORRT_VER=$(cat /etc/TENSORRT_VER) /deps/build_python_tensorrt.sh
 
 COPY docker/tensorrt/requirements-arm64.txt /requirements-tensorrt.txt
-ADD https://nvidia.box.com/shared/static/9aemm4grzbbkfaesg5l7fplgjtmswhj8.whl /tmp/onnxruntime_gpu-1.15.1-cp39-cp39-linux_aarch64.whl
+ADD https://nvidia.box.com/shared/static/psl23iw3bh7hlgku0mjo1xekxpego3e3.whl /tmp/onnxruntime_gpu-1.15.1-cp311-cp311-linux_aarch64.whl
 
 RUN pip3 uninstall -y onnxruntime-openvino \
     && pip3 wheel --wheel-dir=/trt-wheels -r /requirements-tensorrt.txt \
-    && pip3 install --no-deps /tmp/onnxruntime_gpu-1.15.1-cp39-cp39-linux_aarch64.whl
+    && pip3 install --no-deps /tmp/onnxruntime_gpu-1.15.1-cp311-cp311-linux_aarch64.whl
 
 FROM build-wheels AS trt-model-wheels
 ARG DEBIAN_FRONTEND

docker/tensorrt/Dockerfile.base

@@ -3,7 +3,7 @@
 # https://askubuntu.com/questions/972516/debian-frontend-environment-variable
 ARG DEBIAN_FRONTEND=noninteractive
 
-ARG TRT_BASE=nvcr.io/nvidia/tensorrt:23.03-py3
+ARG TRT_BASE=nvcr.io/nvidia/tensorrt:23.12-py3
 
 # Build TensorRT-specific library
 FROM ${TRT_BASE} AS trt-deps
@@ -24,6 +24,7 @@ ENV S6_CMD_WAIT_FOR_SERVICES_MAXTIME=0
 
 COPY --from=trt-deps /usr/local/lib/libyolo_layer.so /usr/local/lib/libyolo_layer.so
 COPY --from=trt-deps /usr/local/src/tensorrt_demos /usr/local/src/tensorrt_demos
+COPY --from=trt-deps /usr/local/cuda-12.* /usr/local/cuda
 COPY docker/tensorrt/detector/rootfs/ /
 ENV YOLO_MODELS=""
 

docker/tensorrt/detector/rootfs/etc/ld.so.conf.d/cuda_tensorrt.conf

@@ -1,6 +1,8 @@
 /usr/local/lib
-/usr/local/lib/python3.9/dist-packages/nvidia/cudnn/lib
-/usr/local/lib/python3.9/dist-packages/nvidia/cuda_runtime/lib
-/usr/local/lib/python3.9/dist-packages/nvidia/cublas/lib
-/usr/local/lib/python3.9/dist-packages/nvidia/cuda_nvrtc/lib
-/usr/local/lib/python3.9/dist-packages/tensorrt
+/usr/local/cuda/lib64
+/usr/local/lib/python3.11/dist-packages/nvidia/cudnn/lib
+/usr/local/lib/python3.11/dist-packages/nvidia/cuda_runtime/lib
+/usr/local/lib/python3.11/dist-packages/nvidia/cublas/lib
+/usr/local/lib/python3.11/dist-packages/nvidia/cuda_nvrtc/lib
+/usr/local/lib/python3.11/dist-packages/tensorrt
+/usr/local/lib/python3.11/dist-packages/nvidia/cufft/lib

docker/tensorrt/requirements-amd64.txt

@@ -1,9 +1,9 @@
 # NVidia TensorRT Support (amd64 only)
 --extra-index-url 'https://pypi.nvidia.com'
 numpy < 1.24; platform_machine == 'x86_64'
-tensorrt == 8.5.3.*; platform_machine == 'x86_64'
-cuda-python == 11.8; platform_machine == 'x86_64'
-cython == 0.29.*; platform_machine == 'x86_64'
+tensorrt == 8.6.1.*; platform_machine == 'x86_64'
+cuda-python == 11.8.*; platform_machine == 'x86_64'
+cython == 3.0.*; platform_machine == 'x86_64'
 nvidia-cuda-runtime-cu12 == 12.1.*; platform_machine == 'x86_64'
 nvidia-cuda-runtime-cu11 == 11.8.*; platform_machine == 'x86_64'
 nvidia-cublas-cu11 == 11.11.3.6; platform_machine == 'x86_64'

docs/docs/configuration/autotracking.md

@@ -41,6 +41,7 @@ cameras:
     ...
     onvif:
       # Required: host of the camera being connected to.
+      # NOTE: HTTP is assumed by default; HTTPS is supported if you specify the scheme, ex: "https://0.0.0.0".
       host: 0.0.0.0
       # Optional: ONVIF port for device (default: shown below).
       port: 8000
@@ -49,6 +50,8 @@ cameras:
       user: admin
       # Optional: password for login.
       password: admin
+      # Optional: Skip TLS verification from the ONVIF server (default: shown below)
+      tls_insecure: False
       # Optional: PTZ camera object autotracking. Keeps a moving object in
       # the center of the frame by automatically moving the PTZ camera.
       autotracking:

docs/docs/configuration/camera_specific.md

@@ -67,14 +67,15 @@ ffmpeg:
 
 ### Annke C800
 
-This camera is H.265 only. To be able to play clips on some devices (like MacOs or iPhone) the H.265 stream has to be repackaged and the audio stream has to be converted to aac. Unfortunately direct playback of in the browser is not working (yet), but the downloaded clip can be played locally.
+This camera is H.265 only. To be able to play clips on some devices (like MacOs or iPhone) the H.265 stream has to be adjusted using the `apple_compatibility` config.
 
 ```yaml
 cameras:
   annkec800: # <------ Name the camera
     ffmpeg:
+      apple_compatibility: true  # <- Adds compatibility with MacOS and iPhone
       output_args:
-        record: -f segment -segment_time 10 -segment_format mp4 -reset_timestamps 1 -strftime 1 -c:v copy -tag:v hvc1 -bsf:v hevc_mp4toannexb -c:a aac
+        record: preset-record-generic-audio-aac
 
       inputs:
         - path: rtsp://user:password@camera-ip:554/H264/ch1/main/av_stream # <----- Update for your camera
@@ -156,7 +157,9 @@ cameras:
 
 #### Reolink Doorbell
 
-The reolink doorbell supports 2-way audio via go2rtc and other applications. It is important that the http-flv stream is still used for stability, a secondary rtsp stream can be added that will be using for the two way audio only.
+The reolink doorbell supports two way audio via go2rtc and other applications. It is important that the http-flv stream is still used for stability, a secondary rtsp stream can be added that will be using for the two way audio only.
+
+Ensure HTTP is enabled in the camera's advanced network settings. To use two way talk with Frigate, see the [Live view documentation](/configuration/live#two-way-talk).
 
 ```yaml
 go2rtc:

docs/docs/configuration/face_recognition.md

@@ -0,0 +1,35 @@
+---
+id: face_recognition
+title: Face Recognition
+---
+
+Face recognition allows people to be assigned names and when their face is recognized Frigate will assign the person's name as a sub label. This information is included in the UI, filters, as well as in notifications.
+
+Frigate has support for FaceNet to create face embeddings, which runs locally. Embeddings are then saved to Frigate's database.
+
+## Minimum System Requirements
+
+Face recognition works by running a large AI model locally on your system. Systems without a GPU will not run Face Recognition reliably or at all.
+
+## Configuration
+
+Face recognition is disabled by default and requires semantic search to be enabled, face recognition must be enabled in your config file before it can be used. Semantic Search and face recognition are global configuration settings.
+
+```yaml
+face_recognition:
+  enabled: true
+```
+
+## Dataset
+
+The number of images needed for a sufficient training set for face recognition varies depending on several factors:
+
+- Complexity of the task: A simple task like recognizing faces of known individuals may require fewer images than a complex task like identifying unknown individuals in a large crowd.
+- Diversity of the dataset: A dataset with diverse images, including variations in lighting, pose, and facial expressions, will require fewer images per person than a less diverse dataset.
+- Desired accuracy: The higher the desired accuracy, the more images are typically needed.
+
+However, here are some general guidelines:
+
+- Minimum: For basic face recognition tasks, a minimum of 10-20 images per person is often recommended.
+- Recommended: For more robust and accurate systems, 30-50 images per person is a good starting point.
+- Ideal: For optimal performance, especially in challenging conditions, 100 or more images per person can be beneficial.

docs/docs/configuration/genai.md

@@ -5,6 +5,8 @@ title: Generative AI
 
 Generative AI can be used to automatically generate descriptive text based on the thumbnails of your tracked objects. This helps with [Semantic Search](/configuration/semantic_search) in Frigate to provide more context about your tracked objects. Descriptions are accessed via the _Explore_ view in the Frigate UI by clicking on a tracked object's thumbnail.
 
+Requests for a description are sent off automatically to your AI provider at the end of the tracked object's lifecycle. Descriptions can also be regenerated manually via the Frigate UI.
+
 :::info
 
 Semantic Search must be enabled to use Generative AI.

docs/docs/configuration/hardware_acceleration.md

@@ -175,6 +175,16 @@ For more information on the various values across different distributions, see h
 
 Depending on your OS and kernel configuration, you may need to change the `/proc/sys/kernel/perf_event_paranoid` kernel tunable. You can test the change by running `sudo sh -c 'echo 2 >/proc/sys/kernel/perf_event_paranoid'` which will persist until a reboot. Make it permanent by running `sudo sh -c 'echo kernel.perf_event_paranoid=2 >> /etc/sysctl.d/local.conf'`
 
+#### Stats for SR-IOV devices
+
+When using virtualized GPUs via SR-IOV, additional args are needed for GPU stats to function. This can be enabled with the following config:
+
+```yaml
+telemetry:
+  stats:
+    sriov: True
+```
+
 ## AMD/ATI GPUs (Radeon HD 2000 and newer GPUs) via libva-mesa-driver
 
 VAAPI supports automatic profile selection so it will work automatically with both H.264 and H.265 streams.

docs/docs/configuration/index.md

@@ -203,14 +203,13 @@ detectors:
   ov:
     type: openvino
     device: AUTO
-    model:
-      path: /openvino-model/ssdlite_mobilenet_v2.xml
 
 model:
   width: 300
   height: 300
   input_tensor: nhwc
   input_pixel_format: bgr
+  path: /openvino-model/ssdlite_mobilenet_v2.xml
   labelmap_path: /openvino-model/coco_91cl_bkgr.txt
 
 record:

docs/docs/configuration/license_plate_recognition.md

@@ -0,0 +1,45 @@
+---
+id: license_plate_recognition
+title: License Plate Recognition (LPR)
+---
+
+Frigate can recognize license plates on vehicles and automatically add the detected characters as a `sub_label` to objects that are of type `car`. A common use case may be to read the license plates of cars pulling into a driveway or cars passing by on a street with a dedicated LPR camera.
+
+Users running a Frigate+ model should ensure that `license_plate` is added to the [list of objects to track](https://docs.frigate.video/plus/#available-label-types) either globally or for a specific camera. This will improve the accuracy and performance of the LPR model.
+
+LPR is most effective when the vehicle’s license plate is fully visible to the camera. For moving vehicles, Frigate will attempt to read the plate continuously, refining its detection and keeping the most confident result. LPR will not run on stationary vehicles.
+
+## Minimum System Requirements
+
+License plate recognition works by running AI models locally on your system. The models are relatively lightweight and run on your CPU. At least 4GB of RAM is required.
+
+## Configuration
+
+License plate recognition is disabled by default. Enable it in your config file:
+
+```yaml
+lpr:
+  enabled: true
+```
+
+## Advanced Configuration
+
+Several options are available to fine-tune the LPR feature. For example, you can adjust the `min_area` setting, which defines the minimum size in pixels a license plate must be before LPR runs. The default is 500 pixels.
+
+Additionally, you can define `known_plates` as strings or regular expressions, allowing Frigate to label tracked vehicles with custom sub_labels when a recognized plate is detected. This information is then accessible in the UI, filters, and notifications.
+
+```yaml
+lpr:
+  enabled: true
+  min_area: 500
+  known_plates:
+    Wife's Car:
+      - "ABC-1234"
+      - "ABC-I234"
+    Johnny:
+      - "J*N-*234" # Using wildcards for H/M and 1/I
+    Sally:
+      - "[S5]LL-1234" # Matches SLL-1234 and 5LL-1234
+```
+
+In this example, "Wife's Car" will appear as the label for any vehicle matching the plate "ABC-1234." The model might occasionally interpret the digit 1 as a capital I (e.g., "ABC-I234"), so both variations are listed. Similarly, multiple possible variations are specified for Johnny and Sally.

docs/docs/configuration/live.md

@@ -29,7 +29,7 @@ The default video and audio codec on your camera may not always be compatible wi
 
 ### Audio Support
 
-MSE Requires AAC audio, WebRTC requires PCMU/PCMA, or opus audio. If you want to support both MSE and WebRTC then your restream config needs to make sure both are enabled.
+MSE Requires PCMA/PCMU or AAC audio, WebRTC requires PCMA/PCMU or opus audio. If you want to support both MSE and WebRTC then your restream config needs to make sure both are enabled.
 
 ```yaml
 go2rtc:
@@ -138,3 +138,13 @@ services:
 :::
 
 See [go2rtc WebRTC docs](https://github.com/AlexxIT/go2rtc/tree/v1.8.3#module-webrtc) for more information about this.
+
+### Two way talk
+
+For devices that support two way talk, Frigate can be configured to use the feature from the camera's Live view in the Web UI. You should:
+
+- Set up go2rtc with [WebRTC](#webrtc-extra-configuration).
+- Ensure you access Frigate via https (may require [opening port 8971](/frigate/installation/#ports)).
+- For the Home Assistant Frigate card, [follow the docs](https://github.com/dermotduffy/frigate-hass-card?tab=readme-ov-file#using-2-way-audio) for the correct source.
+
+To use the Reolink Doorbell with two way talk, you should use the [recommended Reolink configuration](/configuration/camera_specific#reolink-doorbell)

docs/docs/configuration/object_detectors.md

@@ -144,7 +144,9 @@ detectors:
 
 #### SSDLite MobileNet v2
 
-An OpenVINO model is provided in the container at `/openvino-model/ssdlite_mobilenet_v2.xml` and is used by this detector type by default. The model comes from Intel's Open Model Zoo [SSDLite MobileNet V2](https://github.com/openvinotoolkit/open_model_zoo/tree/master/models/public/ssdlite_mobilenet_v2) and is converted to an FP16 precision IR model. Use the model configuration shown below when using the OpenVINO detector with the default model.
+An OpenVINO model is provided in the container at `/openvino-model/ssdlite_mobilenet_v2.xml` and is used by this detector type by default. The model comes from Intel's Open Model Zoo [SSDLite MobileNet V2](https://github.com/openvinotoolkit/open_model_zoo/tree/master/models/public/ssdlite_mobilenet_v2) and is converted to an FP16 precision IR model.
+
+Use the model configuration shown below when using the OpenVINO detector with the default OpenVINO model:
 
 ```yaml
 detectors:
@@ -254,6 +256,7 @@ yolov4x-mish-640
 yolov7-tiny-288
 yolov7-tiny-416
 yolov7-640
+yolov7-416
 yolov7-320
 yolov7x-640
 yolov7x-320
@@ -282,6 +285,8 @@ The TensorRT detector can be selected by specifying `tensorrt` as the model type
 
 The TensorRT detector uses `.trt` model files that are located in `/config/model_cache/tensorrt` by default. These model path and dimensions used will depend on which model you have generated.
 
+Use the config below to work with generated TRT models:
+
 ```yaml
 detectors:
   tensorrt:
@@ -501,11 +506,12 @@ detectors:
   cpu1:
     type: cpu
     num_threads: 3
-    model:
-      path: "/custom_model.tflite"
   cpu2:
     type: cpu
     num_threads: 3
+
+model:
+  path: "/custom_model.tflite"
 ```
 
 When using CPU detectors, you can add one CPU detector per camera. Adding more detectors than the number of cameras should not improve performance.
@@ -544,7 +550,7 @@ Hardware accelerated object detection is supported on the following SoCs:
 - RK3576
 - RK3588
 
-This implementation uses the [Rockchip's RKNN-Toolkit2](https://github.com/airockchip/rknn-toolkit2/), version v2.0.0.beta0. Currently, only [Yolo-NAS](https://github.com/Deci-AI/super-gradients/blob/master/YOLONAS.md) is supported as object detection model.
+This implementation uses the [Rockchip's RKNN-Toolkit2](https://github.com/airockchip/rknn-toolkit2/), version v2.3.0. Currently, only [Yolo-NAS](https://github.com/Deci-AI/super-gradients/blob/master/YOLONAS.md) is supported as object detection model.
 
 ### Prerequisites
 
@@ -617,7 +623,41 @@ $ cat /sys/kernel/debug/rknpu/load
 :::
 
 - All models are automatically downloaded and stored in the folder `config/model_cache/rknn_cache`. After upgrading Frigate, you should remove older models to free up space.
-- You can also provide your own `.rknn` model. You should not save your own models in the `rknn_cache` folder, store them directly in the `model_cache` folder or another subfolder. To convert a model to `.rknn` format see the `rknn-toolkit2` (requires a x86 machine). Note, that there is only post-processing for the supported models.
+- You can also provide your own `.rknn` model. You should not save your own models in the `rknn_cache` folder, store them directly in the `model_cache` folder or another subfolder. To convert a model to `.rknn` format see the `rknn-toolkit2`. Note, that there is only post-processing for the supported models.
+
+### Converting your own onnx model to rknn format
+
+To convert a onnx model to the rknn format using the [rknn-toolkit2](https://github.com/airockchip/rknn-toolkit2/) you have to:
+
+- Place one ore more models in onnx format in the directory `config/model_cache/rknn_cache/onnx` on your docker host (this might require `sudo` privileges).
+- Save the configuration file under `config/conv2rknn.yaml` (see below for details).
+- Run `docker exec <frigate_container_id> python3 /opt/conv2rknn.py`. If the conversion was successful, the rknn models will be placed in `config/model_cache/rknn_cache`.
+
+This is an example configuration file that you need to adjust to your specific onnx model:
+
+```yaml
+soc: ["rk3562","rk3566", "rk3568", "rk3576", "rk3588"]
+quantization: false
+
+output_name: "{input_basename}"
+
+config:
+  mean_values: [[0, 0, 0]]
+  std_values: [[255, 255, 255]]
+  quant_img_rgb2bgr: true
+```
+
+Explanation of the paramters:
+
+- `soc`: A list of all SoCs you want to build the rknn model for. If you don't specify this parameter, the script tries to find out your SoC and builds the rknn model for this one.
+- `quantization`: true: 8 bit integer (i8) quantization, false: 16 bit float (fp16). Default: false.
+- `output_name`: The output name of the model. The following variables are available:
+  - `quant`: "i8" or "fp16" depending on the config
+  - `input_basename`: the basename of the input model (e.g. "my_model" if the input model is calles "my_model.onnx")
+  - `soc`: the SoC this model was build for (e.g. "rk3588")
+  - `tk_version`: Version of `rknn-toolkit2` (e.g. "2.3.0")
+  - **example**: Specifying `output_name = "frigate-{quant}-{input_basename}-{soc}-v{tk_version}"` could result in a model called `frigate-i8-my_model-rk3588-v2.3.0.rknn`.
+- `config`: Configuration passed to `rknn-toolkit2` for model conversion. For an explanation of all available parameters have a look at section "2.2. Model configuration" of [this manual](https://github.com/MarcA711/rknn-toolkit2/releases/download/v2.3.0/03_Rockchip_RKNPU_API_Reference_RKNN_Toolkit2_V2.3.0_EN.pdf).
 
 ## Hailo-8l
 
@@ -632,8 +672,6 @@ detectors:
   hailo8l:
     type: hailo8l
     device: PCIe
-    model:
-      path: /config/model_cache/h8l_cache/ssd_mobilenet_v1.hef
 
 model:
   width: 300
@@ -641,4 +679,5 @@ model:
   input_tensor: nhwc
   input_pixel_format: bgr
   model_type: ssd
+  path: /config/model_cache/h8l_cache/ssd_mobilenet_v1.hef
 ```

docs/docs/configuration/reference.md

@@ -52,7 +52,7 @@ detectors:
   # Required: name of the detector
   detector_name:
     # Required: type of the detector
-    # Frigate provided types include 'cpu', 'edgetpu', 'openvino' and 'tensorrt' (default: shown below)
+    # Frigate provides many types, see https://docs.frigate.video/configuration/object_detectors for more details (default: shown below)
     # Additional detector types can also be plugged in.
     # Detectors may require additional configuration.
     # Refer to the Detectors configuration page for more information.
@@ -117,25 +117,27 @@ auth:
   hash_iterations: 600000
 
 # Optional: model modifications
+# NOTE: The default values are for the EdgeTPU detector.
+# Other detectors will require the model config to be set.
 model:
-  # Optional: path to the model (default: automatic based on detector)
+  # Required: path to the model (default: automatic based on detector)
   path: /edgetpu_model.tflite
-  # Optional: path to the labelmap (default: shown below)
+  # Required: path to the labelmap (default: shown below)
   labelmap_path: /labelmap.txt
   # Required: Object detection model input width (default: shown below)
   width: 320
   # Required: Object detection model input height (default: shown below)
   height: 320
-  # Optional: Object detection model input colorspace
+  # Required: Object detection model input colorspace
   # Valid values are rgb, bgr, or yuv. (default: shown below)
   input_pixel_format: rgb
-  # Optional: Object detection model input tensor format
+  # Required: Object detection model input tensor format
   # Valid values are nhwc or nchw (default: shown below)
   input_tensor: nhwc
-  # Optional: Object detection model type, currently only used with the OpenVINO detector
+  # Required: Object detection model type, currently only used with the OpenVINO detector
   # Valid values are ssd, yolox, yolonas (default: shown below)
   model_type: ssd
-  # Optional: Label name modifications. These are merged into the standard labelmap.
+  # Required: Label name modifications. These are merged into the standard labelmap.
   labelmap:
     2: vehicle
   # Optional: Map of object labels to their attribute labels (default: depends on model)
@@ -242,6 +244,8 @@ ffmpeg:
   # If set too high, then if a ffmpeg crash or camera stream timeout occurs, you could potentially lose up to a maximum of retry_interval second(s) of footage
   # NOTE: this can be a useful setting for Wireless / Battery cameras to reduce how much footage is potentially lost during a connection timeout.
   retry_interval: 10
+  # Optional: Set tag on HEVC (H.265) recording stream to improve compatibility with Apple players. (default: shown below)
+  apple_compatibility: false
 
 # Optional: Detect configuration
 # NOTE: Can be overridden at the camera level
@@ -522,6 +526,14 @@ semantic_search:
   # NOTE: small model runs on CPU and large model runs on GPU
   model_size: "small"
 
+# Optional: Configuration for face recognition capability
+face_recognition:
+  # Optional: Enable semantic search (default: shown below)
+  enabled: False
+  # Optional: Set the model size used for embeddings. (default: shown below)
+  # NOTE: small model runs on CPU and large model runs on GPU
+  model_size: "small"
+
 # Optional: Configuration for AI generated tracked object descriptions
 # NOTE: Semantic Search must be enabled for this to do anything.
 # WARNING: Depending on the provider, this will send thumbnails over the internet
@@ -686,6 +698,7 @@ cameras:
     # to enable PTZ controls.
     onvif:
       # Required: host of the camera being connected to.
+      # NOTE: HTTP is assumed by default; HTTPS is supported if you specify the scheme, ex: "https://0.0.0.0".
       host: 0.0.0.0
       # Optional: ONVIF port for device (default: shown below).
       port: 8000
@@ -694,6 +707,8 @@ cameras:
       user: admin
       # Optional: password for login.
       password: admin
+      # Optional: Skip TLS verification from the ONVIF server (default: shown below)
+      tls_insecure: False
       # Optional: Ignores time synchronization mismatches between the camera and the server during authentication.
       # Using NTP on both ends is recommended and this should only be set to True in a "safe" environment due to the security risk it represents.
       ignore_time_mismatch: False
@@ -757,6 +772,8 @@ cameras:
         - cat
       # Optional: Restrict generation to objects that entered any of the listed zones (default: none, all zones qualify)
       required_zones: []
+      # Optional: Save thumbnails sent to generative AI for review/debugging purposes (default: shown below)
+      debug_save_thumbnails: False
 
 # Optional
 ui:
@@ -798,11 +815,13 @@ telemetry:
     - lo
   # Optional: Configure system stats
   stats:
-    # Enable AMD GPU stats (default: shown below)
+    # Optional: Enable AMD GPU stats (default: shown below)
     amd_gpu_stats: True
-    # Enable Intel GPU stats (default: shown below)
+    # Optional: Enable Intel GPU stats (default: shown below)
     intel_gpu_stats: True
-    # Enable network bandwidth stats monitoring for camera ffmpeg processes, go2rtc, and object detectors. (default: shown below)
+    # Optional: Treat GPU as SR-IOV to fix GPU stats (default: shown below)
+    sriov: False
+    # Optional: Enable network bandwidth stats monitoring for camera ffmpeg processes, go2rtc, and object detectors. (default: shown below)
     # NOTE: The container must either be privileged or have cap_net_admin, cap_net_raw capabilities enabled.
     network_bandwidth: False
   # Optional: Enable the latest version outbound check (default: shown below)

docs/docs/frigate/installation.md

@@ -305,8 +305,15 @@ To install make sure you have the [community app plugin here](https://forums.unr
 
 ## Proxmox
 
-It is recommended to run Frigate in LXC, rather than in a VM, for maximum performance. The setup can be complex so be prepared to read the Proxmox and LXC documentation. Suggestions include:
+[According to Proxmox documentation](https://pve.proxmox.com/pve-docs/pve-admin-guide.html#chapter_pct) it is recommended that you run application containers like Frigate inside a Proxmox QEMU VM. This will give you all the advantages of application containerization, while also providing the benefits that VMs offer, such as strong isolation from the host and the ability to live-migrate, which otherwise isn’t possible with containers.
 
+:::warning
+
+If you choose to run Frigate via LXC in Proxmox the setup can be complex so be prepared to read the Proxmox and LXC documentation, Frigate does not officially support running inside of an LXC.
+
+:::
+
+ Suggestions include:
 - For Intel-based hardware acceleration, to allow access to the `/dev/dri/renderD128` device with major number 226 and minor number 128, add the following lines to the `/etc/pve/lxc/<id>.conf` LXC configuration:
   - `lxc.cgroup2.devices.allow: c 226:128 rwm`
   - `lxc.mount.entry: /dev/dri/renderD128 dev/dri/renderD128 none bind,optional,create=file`

docs/docs/troubleshooting/recordings.md

@@ -3,7 +3,15 @@ id: recordings
 title: Troubleshooting Recordings
 ---
 
-### WARNING : Unable to keep up with recording segments in cache for camera. Keeping the 5 most recent segments out of 6 and discarding the rest...
+## I have Frigate configured for motion recording only, but it still seems to be recording even with no motion. Why?
+
+You'll want to:
+
+- Make sure your camera's timestamp is masked out with a motion mask. Even if there is no motion occurring in your scene, your motion settings may be sensitive enough to count your timestamp as motion.
+- If you have audio detection enabled, keep in mind that audio that is heard above `min_volume` is considered motion.
+- [Tune your motion detection settings](/configuration/motion_detection) either by editing your config file or by using the UI's Motion Tuner.
+
+## I see the message: WARNING : Unable to keep up with recording segments in cache for camera. Keeping the 5 most recent segments out of 6 and discarding the rest...
 
 This error can be caused by a number of different issues. The first step in troubleshooting is to enable debug logging for recording. This will enable logging showing how long it takes for recordings to be moved from RAM cache to the disk.
 
@@ -40,6 +48,7 @@ On linux, some helpful tools/commands in diagnosing would be:
 On modern linux kernels, the system will utilize some swap if enabled. Setting vm.swappiness=1 no longer means that the kernel will only swap in order to avoid OOM. To prevent any swapping inside a container, set allocations memory and memory+swap to be the same and disable swapping by setting the following docker/podman run parameters:
 
 **Compose example**
+
 ```yaml
 version: "3.9"
 services:
@@ -54,6 +63,7 @@ services:
 ```
 
 **Run command example**
+
 ```
 --memory=<MAXRAM> --memory-swap=<MAXSWAP> --memory-swappiness=0
 ```

docs/package.json

@@ -17,15 +17,15 @@
     "write-heading-ids": "docusaurus write-heading-ids"
   },
   "dependencies": {
-    "@docusaurus/core": "^3.5.2",
-    "@docusaurus/preset-classic": "^3.5.2",
-    "@docusaurus/theme-mermaid": "^3.5.2",
-    "@docusaurus/plugin-content-docs": "^3.5.2",
-    "@mdx-js/react": "^3.0.1",
+    "@docusaurus/core": "^3.6.3",
+    "@docusaurus/preset-classic": "^3.6.3",
+    "@docusaurus/theme-mermaid": "^3.6.3",
+    "@docusaurus/plugin-content-docs": "^3.6.3",
+    "@mdx-js/react": "^3.1.0",
     "clsx": "^2.1.1",
-    "docusaurus-plugin-openapi-docs": "^4.1.0",
-    "docusaurus-theme-openapi-docs": "^4.1.0",
-    "prism-react-renderer": "^2.4.0",
+    "docusaurus-plugin-openapi-docs": "^4.3.1",
+    "docusaurus-theme-openapi-docs": "^4.3.1",
+    "prism-react-renderer": "^2.4.1",
     "raw-loader": "^4.0.2",
     "react": "^18.3.1",
     "react-dom": "^18.3.1"

docs/sidebars.ts

@@ -36,6 +36,8 @@ const sidebars: SidebarsConfig = {
       'Semantic Search': [
         'configuration/semantic_search',
         'configuration/genai',
+        'configuration/face_recognition',
+        'configuration/license_plate_recognition',
       ],
       Cameras: [
         'configuration/cameras',

frigate/__main__.py

@@ -3,12 +3,15 @@ import faulthandler
 import signal
 import sys
 import threading
+from typing import Union
 
+import ruamel.yaml
 from pydantic import ValidationError
 
 from frigate.app import FrigateApp
 from frigate.config import FrigateConfig
 from frigate.log import setup_logging
+from frigate.util.config import find_config_file
 
 
 def main() -> None:
@@ -42,10 +45,51 @@ def main() -> None:
         print("*************************************************************")
         print("*************************************************************")
         print("***    Config Validation Errors                           ***")
-        print("*************************************************************")
+        print("*************************************************************\n")
+        # Attempt to get the original config file for line number tracking
+        config_path = find_config_file()
+        with open(config_path, "r") as f:
+            yaml_config = ruamel.yaml.YAML()
+            yaml_config.preserve_quotes = True
+            full_config = yaml_config.load(f)
+
         for error in e.errors():
-            location = ".".join(str(item) for item in error["loc"])
-            print(f"{location}: {error['msg']}")
+            error_path = error["loc"]
+
+            current = full_config
+            line_number = "Unknown"
+            last_line_number = "Unknown"
+
+            try:
+                for i, part in enumerate(error_path):
+                    key: Union[int, str] = (
+                        int(part) if isinstance(part, str) and part.isdigit() else part
+                    )
+
+                    if isinstance(current, ruamel.yaml.comments.CommentedMap):
+                        current = current[key]
+                    elif isinstance(current, list):
+                        if isinstance(key, int):
+                            current = current[key]
+
+                    if hasattr(current, "lc"):
+                        last_line_number = current.lc.line
+
+                    if i == len(error_path) - 1:
+                        if hasattr(current, "lc"):
+                            line_number = current.lc.line
+                        else:
+                            line_number = last_line_number
+
+            except Exception as traverse_error:
+                print(f"Could not determine exact line number: {traverse_error}")
+
+            if current != full_config:
+                print(f"Line #  : {line_number}")
+                print(f"Key     : {' -> '.join(map(str, error_path))}")
+                print(f"Value   : {error.get('input', '-')}")
+            print(f"Message : {error.get('msg', error.get('type', 'Unknown'))}\n")
+
         print("*************************************************************")
         print("***    End Config Validation Errors                       ***")
         print("*************************************************************")

frigate/api/app.py

@@ -7,15 +7,18 @@ import os
 import traceback
 from datetime import datetime, timedelta
 from functools import reduce
+from io import StringIO
 from typing import Any, Optional
 
 import requests
+import ruamel.yaml
 from fastapi import APIRouter, Body, Path, Request, Response
 from fastapi.encoders import jsonable_encoder
 from fastapi.params import Depends
 from fastapi.responses import JSONResponse, PlainTextResponse
 from markupsafe import escape
 from peewee import operator
+from pydantic import ValidationError
 
 from frigate.api.defs.query.app_query_parameters import AppTimelineHourlyQueryParameters
 from frigate.api.defs.request.app_body import AppConfigSetBody
@@ -139,6 +142,8 @@ def config(request: Request):
         mode="json", warnings="none", exclude_none=True
     )
     for stream_name, stream in go2rtc.get("streams", {}).items():
+        if stream is None:
+            continue
         if isinstance(stream, str):
             cleaned = clean_camera_user_pass(stream)
         else:
@@ -183,7 +188,6 @@ def config_raw():
 @router.post("/config/save")
 def config_save(save_option: str, body: Any = Body(media_type="text/plain")):
     new_config = body.decode()
-
     if not new_config:
         return JSONResponse(
             content=(
@@ -194,13 +198,64 @@ def config_save(save_option: str, body: Any = Body(media_type="text/plain")):
 
     # Validate the config schema
     try:
+        # Use ruamel to parse and preserve line numbers
+        yaml_config = ruamel.yaml.YAML()
+        yaml_config.preserve_quotes = True
+        full_config = yaml_config.load(StringIO(new_config))
+
         FrigateConfig.parse_yaml(new_config)
+
+    except ValidationError as e:
+        error_message = []
+
+        for error in e.errors():
+            error_path = error["loc"]
+            current = full_config
+            line_number = "Unknown"
+            last_line_number = "Unknown"
+
+            try:
+                for i, part in enumerate(error_path):
+                    key = int(part) if part.isdigit() else part
+
+                    if isinstance(current, ruamel.yaml.comments.CommentedMap):
+                        current = current[key]
+                    elif isinstance(current, list):
+                        current = current[key]
+
+                    if hasattr(current, "lc"):
+                        last_line_number = current.lc.line
+
+                    if i == len(error_path) - 1:
+                        if hasattr(current, "lc"):
+                            line_number = current.lc.line
+                        else:
+                            line_number = last_line_number
+
+            except Exception:
+                line_number = "Unable to determine"
+
+            error_message.append(
+                f"Line {line_number}: {' -> '.join(map(str, error_path))} - {error.get('msg', error.get('type', 'Unknown'))}"
+            )
+
+        return JSONResponse(
+            content=(
+                {
+                    "success": False,
+                    "message": "Your configuration is invalid.\nSee the official documentation at docs.frigate.video.\n\n"
+                    + "\n".join(error_message),
+                }
+            ),
+            status_code=400,
+        )
+
     except Exception:
         return JSONResponse(
             content=(
                 {
                     "success": False,
-                    "message": f"\nConfig Error:\n\n{escape(str(traceback.format_exc()))}",
+                    "message": f"\nYour configuration is invalid.\nSee the official documentation at docs.frigate.video.\n\n{escape(str(traceback.format_exc()))}",
                 }
             ),
             status_code=400,

frigate/api/classification.py

@@ -0,0 +1,127 @@
+"""Object classification APIs."""
+
+import logging
+import os
+import random
+import shutil
+import string
+
+from fastapi import APIRouter, Request, UploadFile
+from fastapi.responses import JSONResponse
+from pathvalidate import sanitize_filename
+
+from frigate.api.defs.tags import Tags
+from frigate.const import FACE_DIR
+from frigate.embeddings import EmbeddingsContext
+
+logger = logging.getLogger(__name__)
+
+router = APIRouter(tags=[Tags.events])
+
+
+@router.get("/faces")
+def get_faces():
+    face_dict: dict[str, list[str]] = {}
+
+    for name in os.listdir(FACE_DIR):
+        face_dir = os.path.join(FACE_DIR, name)
+
+        if not os.path.isdir(face_dir):
+            continue
+
+        face_dict[name] = []
+
+        for file in sorted(
+            os.listdir(face_dir),
+            key=lambda f: os.path.getctime(os.path.join(face_dir, f)),
+            reverse=True,
+        ):
+            face_dict[name].append(file)
+
+    return JSONResponse(status_code=200, content=face_dict)
+
+
+@router.post("/faces/{name}")
+async def register_face(request: Request, name: str, file: UploadFile):
+    if not request.app.frigate_config.face_recognition.enabled:
+        return JSONResponse(
+            status_code=400,
+            content={"message": "Face recognition is not enabled.", "success": False},
+        )
+
+    context: EmbeddingsContext = request.app.embeddings
+    result = context.register_face(name, await file.read())
+    return JSONResponse(
+        status_code=200 if result.get("success", True) else 400,
+        content=result,
+    )
+
+
+@router.post("/faces/train/{name}/classify")
+def train_face(request: Request, name: str, body: dict = None):
+    if not request.app.frigate_config.face_recognition.enabled:
+        return JSONResponse(
+            status_code=400,
+            content={"message": "Face recognition is not enabled.", "success": False},
+        )
+
+    json: dict[str, any] = body or {}
+    training_file = os.path.join(
+        FACE_DIR, f"train/{sanitize_filename(json.get('training_file', ''))}"
+    )
+
+    if not training_file or not os.path.isfile(training_file):
+        return JSONResponse(
+            content=(
+                {
+                    "success": False,
+                    "message": f"Invalid filename or no file exists: {training_file}",
+                }
+            ),
+            status_code=404,
+        )
+
+    rand_id = "".join(random.choices(string.ascii_lowercase + string.digits, k=6))
+    new_name = f"{name}-{rand_id}.webp"
+    new_file = os.path.join(FACE_DIR, f"{name}/{new_name}")
+    shutil.move(training_file, new_file)
+
+    context: EmbeddingsContext = request.app.embeddings
+    context.clear_face_classifier()
+
+    return JSONResponse(
+        content=(
+            {
+                "success": True,
+                "message": f"Successfully saved {training_file} as {new_name}.",
+            }
+        ),
+        status_code=200,
+    )
+
+
+@router.post("/faces/{name}/delete")
+def deregister_faces(request: Request, name: str, body: dict = None):
+    if not request.app.frigate_config.face_recognition.enabled:
+        return JSONResponse(
+            status_code=400,
+            content={"message": "Face recognition is not enabled.", "success": False},
+        )
+
+    json: dict[str, any] = body or {}
+    list_of_ids = json.get("ids", "")
+
+    if not list_of_ids or len(list_of_ids) == 0:
+        return JSONResponse(
+            content=({"success": False, "message": "Not a valid list of ids"}),
+            status_code=404,
+        )
+
+    context: EmbeddingsContext = request.app.embeddings
+    context.delete_face_ids(
+        name, map(lambda file: sanitize_filename(file), list_of_ids)
+    )
+    return JSONResponse(
+        content=({"success": True, "message": "Successfully deleted faces."}),
+        status_code=200,
+    )

frigate/api/defs/query/media_query_parameters.py

@@ -20,6 +20,7 @@ class MediaLatestFrameQueryParams(BaseModel):
     regions: Optional[int] = None
     quality: Optional[int] = 70
     height: Optional[int] = None
+    store: Optional[int] = None
 
 
 class MediaEventsSnapshotQueryParams(BaseModel):

frigate/api/defs/request/events_body.py

@@ -8,6 +8,9 @@ class EventsSubLabelBody(BaseModel):
     subLabelScore: Optional[float] = Field(
         title="Score for sub label", default=None, gt=0.0, le=1.0
     )
+    camera: Optional[str] = Field(
+        title="Camera this object is detected on.", default=None
+    )
 
 
 class EventsDescriptionBody(BaseModel):

frigate/api/defs/tags.py

@@ -10,4 +10,5 @@ class Tags(Enum):
     review = "Review"
     export = "Export"
     events = "Events"
+    classification = "classification"
     auth = "Auth"

frigate/api/event.py

@@ -909,38 +909,59 @@ def set_sub_label(
     try:
         event: Event = Event.get(Event.id == event_id)
     except DoesNotExist:
+        if not body.camera:
+            return JSONResponse(
+                content=(
+                    {
+                        "success": False,
+                        "message": "Event "
+                        + event_id
+                        + " not found and camera is not provided.",
+                    }
+                ),
+                status_code=404,
+            )
+
+        event = None
+
+    if request.app.detected_frames_processor:
+        tracked_obj: TrackedObject = (
+            request.app.detected_frames_processor.camera_states[
+                event.camera if event else body.camera
+            ].tracked_objects.get(event_id)
+        )
+    else:
+        tracked_obj = None
+
+    if not event and not tracked_obj:
         return JSONResponse(
-            content=({"success": False, "message": "Event " + event_id + " not found"}),
+            content=(
+                {"success": False, "message": "Event " + event_id + " not found."}
+            ),
             status_code=404,
         )
 
     new_sub_label = body.subLabel
     new_score = body.subLabelScore
 
-    if not event.end_time:
-        # update tracked object
-        tracked_obj: TrackedObject = (
-            request.app.detected_frames_processor.camera_states[
-                event.camera
-            ].tracked_objects.get(event.id)
-        )
-
-        if tracked_obj:
-            tracked_obj.obj_data["sub_label"] = (new_sub_label, new_score)
+    if tracked_obj:
+        tracked_obj.obj_data["sub_label"] = (new_sub_label, new_score)
 
         # update timeline items
         Timeline.update(
             data=Timeline.data.update({"sub_label": (new_sub_label, new_score)})
         ).where(Timeline.source_id == event_id).execute()
 
-    event.sub_label = new_sub_label
+    if event:
+        event.sub_label = new_sub_label
 
-    if new_score:
-        data = event.data
-        data["sub_label_score"] = new_score
-        event.data = data
+        if new_score:
+            data = event.data
+            data["sub_label_score"] = new_score
+            event.data = data
+
+        event.save()
 
-    event.save()
     return JSONResponse(
         content=(
             {

frigate/api/fastapi_app.py

@@ -11,7 +11,16 @@ from starlette_context import middleware, plugins
 from starlette_context.plugins import Plugin
 
 from frigate.api import app as main_app
-from frigate.api import auth, event, export, media, notification, preview, review
+from frigate.api import (
+    auth,
+    classification,
+    event,
+    export,
+    media,
+    notification,
+    preview,
+    review,
+)
 from frigate.api.auth import get_jwt_secret, limiter
 from frigate.comms.event_metadata_updater import (
     EventMetadataPublisher,
@@ -99,6 +108,7 @@ def create_fastapi_app(
     # Routes
     # Order of include_router matters: https://fastapi.tiangolo.com/tutorial/path-params/#order-matters
     app.include_router(auth.router)
+    app.include_router(classification.router)
     app.include_router(review.router)
     app.include_router(main_app.router)
     app.include_router(preview.router)

frigate/api/media.py

@@ -179,7 +179,12 @@ def latest_frame(
         return Response(
             content=img.tobytes(),
             media_type=f"image/{extension}",
-            headers={"Content-Type": f"image/{extension}", "Cache-Control": "no-store"},
+            headers={
+                "Content-Type": f"image/{extension}",
+                "Cache-Control": "no-store"
+                if not params.store
+                else "private, max-age=60",
+            },
         )
     elif camera_name == "birdseye" and request.app.frigate_config.birdseye.restream:
         frame = cv2.cvtColor(
@@ -198,7 +203,12 @@ def latest_frame(
         return Response(
             content=img.tobytes(),
             media_type=f"image/{extension}",
-            headers={"Content-Type": f"image/{extension}", "Cache-Control": "no-store"},
+            headers={
+                "Content-Type": f"image/{extension}",
+                "Cache-Control": "no-store"
+                if not params.store
+                else "private, max-age=60",
+            },
         )
     else:
         return JSONResponse(

frigate/app.py

@@ -34,10 +34,12 @@ from frigate.const import (
     CLIPS_DIR,
     CONFIG_DIR,
     EXPORT_DIR,
+    FACE_DIR,
     MODEL_CACHE_DIR,
     RECORD_DIR,
     SHM_FRAMES_VAR,
 )
+from frigate.data_processing.types import DataProcessorMetrics
 from frigate.db.sqlitevecq import SqliteVecQueueDatabase
 from frigate.embeddings import EmbeddingsContext, manage_embeddings
 from frigate.events.audio import AudioProcessor
@@ -88,6 +90,9 @@ class FrigateApp:
         self.detection_shms: list[mp.shared_memory.SharedMemory] = []
         self.log_queue: Queue = mp.Queue()
         self.camera_metrics: dict[str, CameraMetrics] = {}
+        self.embeddings_metrics: DataProcessorMetrics | None = (
+            DataProcessorMetrics() if config.semantic_search.enabled else None
+        )
         self.ptz_metrics: dict[str, PTZMetrics] = {}
         self.processes: dict[str, int] = {}
         self.embeddings: Optional[EmbeddingsContext] = None
@@ -96,14 +101,19 @@ class FrigateApp:
         self.config = config
 
     def ensure_dirs(self) -> None:
-        for d in [
+        dirs = [
             CONFIG_DIR,
             RECORD_DIR,
             f"{CLIPS_DIR}/cache",
             CACHE_DIR,
             MODEL_CACHE_DIR,
             EXPORT_DIR,
-        ]:
+        ]
+
+        if self.config.face_recognition.enabled:
+            dirs.append(FACE_DIR)
+
+        for d in dirs:
             if not os.path.exists(d) and not os.path.islink(d):
                 logger.info(f"Creating directory: {d}")
                 os.makedirs(d)
@@ -229,7 +239,10 @@ class FrigateApp:
         embedding_process = util.Process(
             target=manage_embeddings,
             name="embeddings_manager",
-            args=(self.config,),
+            args=(
+                self.config,
+                self.embeddings_metrics,
+            ),
         )
         embedding_process.daemon = True
         self.embedding_process = embedding_process
@@ -437,7 +450,7 @@ class FrigateApp:
             # pre-create shms
             for i in range(shm_frame_count):
                 frame_size = config.frame_shape_yuv[0] * config.frame_shape_yuv[1]
-                self.frame_manager.create(f"{config.name}_{i}", frame_size)
+                self.frame_manager.create(f"{config.name}_frame{i}", frame_size)
 
             capture_process = util.Process(
                 target=capture_camera,
@@ -491,7 +504,11 @@ class FrigateApp:
         self.stats_emitter = StatsEmitter(
             self.config,
             stats_init(
-                self.config, self.camera_metrics, self.detectors, self.processes
+                self.config,
+                self.camera_metrics,
+                self.embeddings_metrics,
+                self.detectors,
+                self.processes,
             ),
             self.stop_event,
         )

frigate/camera/activity_manager.py

@@ -0,0 +1,130 @@
+"""Manage camera activity and updating listeners."""
+
+from collections import Counter
+from typing import Callable
+
+from frigate.config.config import FrigateConfig
+
+
+class CameraActivityManager:
+    def __init__(
+        self, config: FrigateConfig, publish: Callable[[str, any], None]
+    ) -> None:
+        self.config = config
+        self.publish = publish
+        self.last_camera_activity: dict[str, dict[str, any]] = {}
+        self.camera_all_object_counts: dict[str, Counter] = {}
+        self.camera_active_object_counts: dict[str, Counter] = {}
+        self.zone_all_object_counts: dict[str, Counter] = {}
+        self.zone_active_object_counts: dict[str, Counter] = {}
+        self.all_zone_labels: dict[str, set[str]] = {}
+
+        for camera_config in config.cameras.values():
+            if not camera_config.enabled:
+                continue
+
+            self.last_camera_activity[camera_config.name] = {}
+            self.camera_all_object_counts[camera_config.name] = Counter()
+            self.camera_active_object_counts[camera_config.name] = Counter()
+
+            for zone, zone_config in camera_config.zones.items():
+                if zone not in self.all_zone_labels:
+                    self.zone_all_object_counts[zone] = Counter()
+                    self.zone_active_object_counts[zone] = Counter()
+                    self.all_zone_labels[zone] = set()
+
+                self.all_zone_labels[zone].update(zone_config.objects)
+
+    def update_activity(self, new_activity: dict[str, dict[str, any]]) -> None:
+        all_objects: list[dict[str, any]] = []
+
+        for camera in new_activity.keys():
+            new_objects = new_activity[camera].get("objects", [])
+            all_objects.extend(new_objects)
+
+            if self.last_camera_activity.get(camera, {}).get("objects") != new_objects:
+                self.compare_camera_activity(camera, new_objects)
+
+        # run through every zone, getting a count of objects in that zone right now
+        for zone, labels in self.all_zone_labels.items():
+            all_zone_objects = Counter(
+                obj["label"].replace("-verified", "")
+                for obj in all_objects
+                if zone in obj["current_zones"]
+            )
+            active_zone_objects = Counter(
+                obj["label"].replace("-verified", "")
+                for obj in all_objects
+                if zone in obj["current_zones"] and not obj["stationary"]
+            )
+            any_changed = False
+
+            # run through each object and check what topics need to be updated for this zone
+            for label in labels:
+                new_count = all_zone_objects[label]
+                new_active_count = active_zone_objects[label]
+
+                if (
+                    new_count != self.zone_all_object_counts[zone][label]
+                    or label not in self.zone_all_object_counts[zone]
+                ):
+                    any_changed = True
+                    self.publish(f"{zone}/{label}", new_count)
+                    self.zone_all_object_counts[zone][label] = new_count
+
+                if (
+                    new_active_count != self.zone_active_object_counts[zone][label]
+                    or label not in self.zone_active_object_counts[zone]
+                ):
+                    any_changed = True
+                    self.publish(f"{zone}/{label}/active", new_active_count)
+                    self.zone_active_object_counts[zone][label] = new_active_count
+
+            if any_changed:
+                self.publish(f"{zone}/all", sum(list(all_zone_objects.values())))
+                self.publish(
+                    f"{zone}/all/active", sum(list(active_zone_objects.values()))
+                )
+
+        self.last_camera_activity = new_activity
+
+    def compare_camera_activity(
+        self, camera: str, new_activity: dict[str, any]
+    ) -> None:
+        all_objects = Counter(
+            obj["label"].replace("-verified", "") for obj in new_activity
+        )
+        active_objects = Counter(
+            obj["label"].replace("-verified", "")
+            for obj in new_activity
+            if not obj["stationary"]
+        )
+        any_changed = False
+
+        # run through each object and check what topics need to be updated
+        for label in self.config.cameras[camera].objects.track:
+            if label in self.config.model.all_attributes:
+                continue
+
+            new_count = all_objects[label]
+            new_active_count = active_objects[label]
+
+            if (
+                new_count != self.camera_all_object_counts[camera][label]
+                or label not in self.camera_all_object_counts[camera]
+            ):
+                any_changed = True
+                self.publish(f"{camera}/{label}", new_count)
+                self.camera_all_object_counts[camera][label] = new_count
+
+            if (
+                new_active_count != self.camera_active_object_counts[camera][label]
+                or label not in self.camera_active_object_counts[camera]
+            ):
+                any_changed = True
+                self.publish(f"{camera}/{label}/active", new_active_count)
+                self.camera_active_object_counts[camera][label] = new_active_count
+
+        if any_changed:
+            self.publish(f"{camera}/all", sum(list(all_objects.values())))
+            self.publish(f"{camera}/all/active", sum(list(active_objects.values())))

frigate/comms/dispatcher.py

@@ -7,6 +7,7 @@ from abc import ABC, abstractmethod
 from typing import Any, Callable, Optional
 
 from frigate.camera import PTZMetrics
+from frigate.camera.activity_manager import CameraActivityManager
 from frigate.comms.config_updater import ConfigPublisher
 from frigate.config import BirdseyeModeEnum, FrigateConfig
 from frigate.const import (
@@ -64,7 +65,7 @@ class Dispatcher:
         self.onvif = onvif
         self.ptz_metrics = ptz_metrics
         self.comms = communicators
-        self.camera_activity = {}
+        self.camera_activity = CameraActivityManager(config, self.publish)
         self.model_state = {}
         self.embeddings_reindex = {}
 
@@ -130,7 +131,7 @@ class Dispatcher:
             ).execute()
 
         def handle_update_camera_activity():
-            self.camera_activity = payload
+            self.camera_activity.update_activity(payload)
 
         def handle_update_event_description():
             event: Event = Event.get(Event.id == payload["id"])
@@ -171,7 +172,7 @@ class Dispatcher:
             )
 
         def handle_on_connect():
-            camera_status = self.camera_activity.copy()
+            camera_status = self.camera_activity.last_camera_activity.copy()
 
             for camera in camera_status.keys():
                 camera_status[camera]["config"] = {

frigate/comms/embeddings_updater.py

@@ -9,9 +9,11 @@ SOCKET_REP_REQ = "ipc:///tmp/cache/embeddings"
 
 
 class EmbeddingsRequestEnum(Enum):
+    clear_face_classifier = "clear_face_classifier"
     embed_description = "embed_description"
     embed_thumbnail = "embed_thumbnail"
     generate_search = "generate_search"
+    register_face = "register_face"
 
 
 class EmbeddingsResponder:
@@ -22,7 +24,7 @@ class EmbeddingsResponder:
 
     def check_for_request(self, process: Callable) -> None:
         while True:  # load all messages that are queued
-            has_message, _, _ = zmq.select([self.socket], [], [], 0.1)
+            has_message, _, _ = zmq.select([self.socket], [], [], 0.01)
 
             if not has_message:
                 break

frigate/comms/webpush.py

@@ -151,7 +151,7 @@ class WebPushClient(Communicator):  # type: ignore[misc]
         camera: str = payload["after"]["camera"]
         title = f"{', '.join(sorted_objects).replace('_', ' ').title()}{' was' if state == 'end' else ''} detected in {', '.join(payload['after']['data']['zones']).replace('_', ' ').title()}"
         message = f"Detected on {camera.replace('_', ' ').title()}"
-        image = f'{payload["after"]["thumb_path"].replace("/media/frigate", "")}'
+        image = f"{payload['after']['thumb_path'].replace('/media/frigate', '')}"
 
         # if event is ongoing open to live view otherwise open to recordings view
         direct_url = f"/review?id={reviewId}" if state == "end" else f"/#{camera}"

frigate/config/__init__.py

@@ -3,13 +3,13 @@ from frigate.detectors import DetectorConfig, ModelConfig  # noqa: F401
 from .auth import *  # noqa: F403
 from .camera import *  # noqa: F403
 from .camera_group import *  # noqa: F403
+from .classification import *  # noqa: F403
 from .config import *  # noqa: F403
 from .database import *  # noqa: F403
 from .logger import *  # noqa: F403
 from .mqtt import *  # noqa: F403
 from .notification import *  # noqa: F403
 from .proxy import *  # noqa: F403
-from .semantic_search import *  # noqa: F403
 from .telemetry import *  # noqa: F403
 from .tls import *  # noqa: F403
 from .ui import *  # noqa: F403

frigate/config/camera/camera.py

@@ -167,7 +167,7 @@ class CameraConfig(FrigateBaseModel):
             record_args = get_ffmpeg_arg_list(
                 parse_preset_output_record(
                     self.ffmpeg.output_args.record,
-                    self.ffmpeg.output_args._force_record_hvc1,
+                    self.ffmpeg.apple_compatibility,
                 )
                 or self.ffmpeg.output_args.record
             )

frigate/config/camera/ffmpeg.py

@@ -2,7 +2,7 @@ import shutil
 from enum import Enum
 from typing import Union
 
-from pydantic import Field, PrivateAttr, field_validator
+from pydantic import Field, field_validator
 
 from frigate.const import DEFAULT_FFMPEG_VERSION, INCLUDED_FFMPEG_VERSIONS
 
@@ -42,7 +42,6 @@ class FfmpegOutputArgsConfig(FrigateBaseModel):
         default=RECORD_FFMPEG_OUTPUT_ARGS_DEFAULT,
         title="Record role FFmpeg output arguments.",
     )
-    _force_record_hvc1: bool = PrivateAttr(default=False)
 
 
 class FfmpegConfig(FrigateBaseModel):
@@ -64,6 +63,10 @@ class FfmpegConfig(FrigateBaseModel):
         default=10.0,
         title="Time in seconds to wait before FFmpeg retries connecting to the camera.",
     )
+    apple_compatibility: bool = Field(
+        default=False,
+        title="Set tag on HEVC (H.265) recording stream to improve compatibility with Apple players.",
+    )
 
     @property
     def ffmpeg_path(self) -> str:

frigate/config/camera/genai.py

@@ -38,6 +38,10 @@ class GenAICameraConfig(BaseModel):
         default_factory=list,
         title="List of required zones to be entered in order to run generative AI.",
     )
+    debug_save_thumbnails: bool = Field(
+        default=False,
+        title="Save thumbnails sent to generative AI for debugging purposes.",
+    )
 
     @field_validator("required_zones", mode="before")
     @classmethod

frigate/config/camera/objects.py

@@ -1,6 +1,6 @@
 from typing import Any, Optional, Union
 
-from pydantic import Field, field_serializer
+from pydantic import Field, PrivateAttr, field_serializer
 
 from ..base import FrigateBaseModel
 
@@ -53,3 +53,20 @@ class ObjectConfig(FrigateBaseModel):
         default_factory=dict, title="Object filters."
     )
     mask: Union[str, list[str]] = Field(default="", title="Object mask.")
+    _all_objects: list[str] = PrivateAttr()
+
+    @property
+    def all_objects(self) -> list[str]:
+        return self._all_objects
+
+    def parse_all_objects(self, cameras):
+        if "_all_objects" in self:
+            return
+
+        # get list of unique enabled labels for tracking
+        enabled_labels = set(self.track)
+
+        for camera in cameras.values():
+            enabled_labels.update(camera.objects.track)
+
+        self._all_objects = list(enabled_labels)

frigate/config/camera/onvif.py

@@ -74,6 +74,7 @@ class OnvifConfig(FrigateBaseModel):
     port: int = Field(default=8000, title="Onvif Port")
     user: Optional[EnvString] = Field(default=None, title="Onvif Username")
     password: Optional[EnvString] = Field(default=None, title="Onvif Password")
+    tls_insecure: bool = Field(default=False, title="Onvif Disable TLS verification")
     autotracking: PtzAutotrackConfig = Field(
         default_factory=PtzAutotrackConfig,
         title="PTZ auto tracking config.",

frigate/config/camera/record.py

@@ -4,6 +4,7 @@ from typing import Optional
 from pydantic import Field
 
 from frigate.const import MAX_PRE_CAPTURE
+from frigate.review.types import SeverityEnum
 
 from ..base import FrigateBaseModel
 
@@ -101,3 +102,15 @@ class RecordConfig(FrigateBaseModel):
             self.alerts.pre_capture,
             self.detections.pre_capture,
         )
+
+    def get_review_pre_capture(self, severity: SeverityEnum) -> int:
+        if severity == SeverityEnum.alert:
+            return self.alerts.pre_capture
+        else:
+            return self.detections.pre_capture
+
+    def get_review_post_capture(self, severity: SeverityEnum) -> int:
+        if severity == SeverityEnum.alert:
+            return self.alerts.post_capture
+        else:
+            return self.detections.post_capture

frigate/config/camera/zone.py

@@ -85,7 +85,7 @@ class ZoneConfig(BaseModel):
             if explicit:
                 self.coordinates = ",".join(
                     [
-                        f'{round(int(p.split(",")[0]) / frame_shape[1], 3)},{round(int(p.split(",")[1]) / frame_shape[0], 3)}'
+                        f"{round(int(p.split(',')[0]) / frame_shape[1], 3)},{round(int(p.split(',')[1]) / frame_shape[0], 3)}"
                         for p in coordinates
                     ]
                 )

frigate/config/classification.py

@@ -0,0 +1,74 @@
+from typing import Dict, List, Optional
+
+from pydantic import Field
+
+from .base import FrigateBaseModel
+
+__all__ = [
+    "FaceRecognitionConfig",
+    "SemanticSearchConfig",
+    "LicensePlateRecognitionConfig",
+]
+
+
+class BirdClassificationConfig(FrigateBaseModel):
+    enabled: bool = Field(default=False, title="Enable bird classification.")
+    threshold: float = Field(
+        default=0.9,
+        title="Minimum classification score required to be considered a match.",
+        gt=0.0,
+        le=1.0,
+    )
+
+
+class ClassificationConfig(FrigateBaseModel):
+    bird: BirdClassificationConfig = Field(
+        default_factory=BirdClassificationConfig, title="Bird classification config."
+    )
+
+
+class SemanticSearchConfig(FrigateBaseModel):
+    enabled: bool = Field(default=False, title="Enable semantic search.")
+    reindex: Optional[bool] = Field(
+        default=False, title="Reindex all detections on startup."
+    )
+    model_size: str = Field(
+        default="small", title="The size of the embeddings model used."
+    )
+
+
+class FaceRecognitionConfig(FrigateBaseModel):
+    enabled: bool = Field(default=False, title="Enable face recognition.")
+    min_score: float = Field(
+        title="Minimum face distance score required to save the attempt.",
+        default=0.8,
+        gt=0.0,
+        le=1.0,
+    )
+    threshold: float = Field(
+        default=0.9,
+        title="Minimum face distance score required to be considered a match.",
+        gt=0.0,
+        le=1.0,
+    )
+    min_area: int = Field(
+        default=500, title="Min area of face box to consider running face recognition."
+    )
+    save_attempts: bool = Field(
+        default=True, title="Save images of face detections for training."
+    )
+
+
+class LicensePlateRecognitionConfig(FrigateBaseModel):
+    enabled: bool = Field(default=False, title="Enable license plate recognition.")
+    threshold: float = Field(
+        default=0.9,
+        title="License plate confidence score required to be added to the object as a sub label.",
+    )
+    min_area: int = Field(
+        default=500,
+        title="Min area of license plate to consider running license plate recognition.",
+    )
+    known_plates: Optional[Dict[str, List[str]]] = Field(
+        default={}, title="Known plates to track."
+    )

frigate/config/config.py

@@ -51,13 +51,18 @@ from .camera.review import ReviewConfig
 from .camera.snapshots import SnapshotsConfig
 from .camera.timestamp import TimestampStyleConfig
 from .camera_group import CameraGroupConfig
+from .classification import (
+    ClassificationConfig,
+    FaceRecognitionConfig,
+    LicensePlateRecognitionConfig,
+    SemanticSearchConfig,
+)
 from .database import DatabaseConfig
 from .env import EnvVars
 from .logger import LoggerConfig
 from .mqtt import MqttConfig
 from .notification import NotificationConfig
 from .proxy import ProxyConfig
-from .semantic_search import SemanticSearchConfig
 from .telemetry import TelemetryConfig
 from .tls import TlsConfig
 from .ui import UIConfig
@@ -159,6 +164,16 @@ class RestreamConfig(BaseModel):
     model_config = ConfigDict(extra="allow")
 
 
+def verify_semantic_search_dependent_configs(config: FrigateConfig) -> None:
+    """Verify that semantic search is enabled if required features are enabled."""
+    if not config.semantic_search.enabled:
+        if config.genai.enabled:
+            raise ValueError("Genai requires semantic search to be enabled.")
+
+        if config.face_recognition.enabled:
+            raise ValueError("Face recognition requires semantic to be enabled.")
+
+
 def verify_config_roles(camera_config: CameraConfig) -> None:
     """Verify that roles are setup in the config correctly."""
     assigned_roles = list(
@@ -317,9 +332,19 @@ class FrigateConfig(FrigateBaseModel):
         default_factory=TelemetryConfig, title="Telemetry configuration."
     )
     tls: TlsConfig = Field(default_factory=TlsConfig, title="TLS configuration.")
+    classification: ClassificationConfig = Field(
+        default_factory=ClassificationConfig, title="Object classification config."
+    )
     semantic_search: SemanticSearchConfig = Field(
         default_factory=SemanticSearchConfig, title="Semantic search configuration."
     )
+    face_recognition: FaceRecognitionConfig = Field(
+        default_factory=FaceRecognitionConfig, title="Face recognition config."
+    )
+    lpr: LicensePlateRecognitionConfig = Field(
+        default_factory=LicensePlateRecognitionConfig,
+        title="License Plate recognition config.",
+    )
     ui: UIConfig = Field(default_factory=UIConfig, title="UI configuration.")
 
     # Detector config
@@ -437,13 +462,12 @@ class FrigateConfig(FrigateBaseModel):
                 camera_config.ffmpeg.hwaccel_args = self.ffmpeg.hwaccel_args
 
             for input in camera_config.ffmpeg.inputs:
-                need_record_fourcc = False and "record" in input.roles
                 need_detect_dimensions = "detect" in input.roles and (
                     camera_config.detect.height is None
                     or camera_config.detect.width is None
                 )
 
-                if need_detect_dimensions or need_record_fourcc:
+                if need_detect_dimensions:
                     stream_info = {"width": 0, "height": 0, "fourcc": None}
                     try:
                         stream_info = stream_info_retriever.get_stream_info(
@@ -467,14 +491,6 @@ class FrigateConfig(FrigateBaseModel):
                         else DEFAULT_DETECT_DIMENSIONS["height"]
                     )
 
-                if need_record_fourcc:
-                    # Apple only supports HEVC if it is hvc1 (vs. hev1)
-                    camera_config.ffmpeg.output_args._force_record_hvc1 = (
-                        stream_info["fourcc"] == "hevc"
-                        if stream_info.get("hevc")
-                        else False
-                    )
-
             # Warn if detect fps > 10
             if camera_config.detect.fps > 10:
                 logger.warning(
@@ -578,13 +594,8 @@ class FrigateConfig(FrigateBaseModel):
             verify_autotrack_zones(camera_config)
             verify_motion_and_detect(camera_config)
 
-        # get list of unique enabled labels for tracking
-        enabled_labels = set(self.objects.track)
-
-        for camera in self.cameras.values():
-            enabled_labels.update(camera.objects.track)
-
-        self.model.create_colormap(sorted(enabled_labels))
+        self.objects.parse_all_objects(self.cameras)
+        self.model.create_colormap(sorted(self.objects.all_objects))
         self.model.check_and_load_plus_model(self.plus_api)
 
         for key, detector in self.detectors.items():
@@ -594,37 +605,30 @@ class FrigateConfig(FrigateBaseModel):
                 if isinstance(detector, dict)
                 else detector.model_dump(warnings="none")
             )
-            detector_config: DetectorConfig = adapter.validate_python(model_dict)
-            if detector_config.model is None:
-                detector_config.model = self.model.model_copy()
-            else:
-                path = detector_config.model.path
-                detector_config.model = self.model.model_copy()
-                detector_config.model.path = path
+            detector_config: BaseDetectorConfig = adapter.validate_python(model_dict)
 
-                if "path" not in model_dict or len(model_dict.keys()) > 1:
-                    logger.warning(
-                        "Customizing more than a detector model path is unsupported."
-                    )
+            # users should not set model themselves
+            if detector_config.model:
+                detector_config.model = None
 
-            merged_model = deep_merge(
-                detector_config.model.model_dump(exclude_unset=True, warnings="none"),
-                self.model.model_dump(exclude_unset=True, warnings="none"),
-            )
+            model_config = self.model.model_dump(exclude_unset=True, warnings="none")
 
-            if "path" not in merged_model:
+            if detector_config.model_path:
+                model_config["path"] = detector_config.model_path
+
+            if "path" not in model_config:
                 if detector_config.type == "cpu":
-                    merged_model["path"] = "/cpu_model.tflite"
+                    model_config["path"] = "/cpu_model.tflite"
                 elif detector_config.type == "edgetpu":
-                    merged_model["path"] = "/edgetpu_model.tflite"
+                    model_config["path"] = "/edgetpu_model.tflite"
 
-            detector_config.model = ModelConfig.model_validate(merged_model)
-            detector_config.model.check_and_load_plus_model(
-                self.plus_api, detector_config.type
-            )
-            detector_config.model.compute_model_hash()
+            model = ModelConfig.model_validate(model_config)
+            model.check_and_load_plus_model(self.plus_api, detector_config.type)
+            model.compute_model_hash()
+            detector_config.model = model
             self.detectors[key] = detector_config
 
+        verify_semantic_search_dependent_configs(self)
         return self
 
     @field_validator("cameras")

frigate/config/logger.py

@@ -29,6 +29,7 @@ class LoggerConfig(FrigateBaseModel):
             logging.getLogger().setLevel(self.default.value.upper())
 
             log_levels = {
+                "httpx": LogLevel.error,
                 "werkzeug": LogLevel.error,
                 "ws4py": LogLevel.error,
                 **self.logs,

frigate/config/semantic_search.py

@@ -1,17 +0,0 @@
-from typing import Optional
-
-from pydantic import Field
-
-from .base import FrigateBaseModel
-
-__all__ = ["SemanticSearchConfig"]
-
-
-class SemanticSearchConfig(FrigateBaseModel):
-    enabled: bool = Field(default=False, title="Enable semantic search.")
-    reindex: Optional[bool] = Field(
-        default=False, title="Reindex all detections on startup."
-    )
-    model_size: str = Field(
-        default="small", title="The size of the embeddings model used."
-    )

frigate/config/telemetry.py

@@ -11,6 +11,9 @@ class StatsConfig(FrigateBaseModel):
     network_bandwidth: bool = Field(
         default=False, title="Enable network bandwidth for ffmpeg processes."
     )
+    sriov: bool = Field(
+        default=False, title="Treat device as SR-IOV to support GPU stats."
+    )
 
 
 class TelemetryConfig(FrigateBaseModel):

frigate/const.py

@@ -5,8 +5,9 @@ DEFAULT_DB_PATH = f"{CONFIG_DIR}/frigate.db"
 MODEL_CACHE_DIR = f"{CONFIG_DIR}/model_cache"
 BASE_DIR = "/media/frigate"
 CLIPS_DIR = f"{BASE_DIR}/clips"
-RECORD_DIR = f"{BASE_DIR}/recordings"
 EXPORT_DIR = f"{BASE_DIR}/exports"
+FACE_DIR = f"{CLIPS_DIR}/faces"
+RECORD_DIR = f"{BASE_DIR}/recordings"
 BIRDSEYE_PIPE = "/tmp/cache/birdseye"
 CACHE_DIR = "/tmp/cache"
 FRIGATE_LOCALHOST = "http://127.0.0.1:5000"
@@ -64,6 +65,7 @@ INCLUDED_FFMPEG_VERSIONS = ["7.0", "5.0"]
 FFMPEG_HWACCEL_NVIDIA = "preset-nvidia"
 FFMPEG_HWACCEL_VAAPI = "preset-vaapi"
 FFMPEG_HWACCEL_VULKAN = "preset-vulkan"
+FFMPEG_HVC1_ARGS = ["-tag:v", "hvc1"]
 
 # Regex constants
 

frigate/data_processing/post/api.py

@@ -0,0 +1,43 @@
+"""Local or remote processors to handle post processing."""
+
+import logging
+from abc import ABC, abstractmethod
+
+from frigate.config import FrigateConfig
+
+from ..types import DataProcessorMetrics, PostProcessDataEnum
+
+logger = logging.getLogger(__name__)
+
+
+class PostProcessorApi(ABC):
+    @abstractmethod
+    def __init__(self, config: FrigateConfig, metrics: DataProcessorMetrics) -> None:
+        self.config = config
+        self.metrics = metrics
+        pass
+
+    @abstractmethod
+    def process_data(
+        self, data: dict[str, any], data_type: PostProcessDataEnum
+    ) -> None:
+        """Processes the data of data type.
+        Args:
+            data (dict): containing data about the input.
+            data_type (enum): Describing the data that is being processed.
+
+        Returns:
+            None.
+        """
+        pass
+
+    @abstractmethod
+    def handle_request(self, request_data: dict[str, any]) -> dict[str, any] | None:
+        """Handle metadata requests.
+        Args:
+            request_data (dict): containing data about requested change to process.
+
+        Returns:
+            None if request was not handled, otherwise return response.
+        """
+        pass

frigate/data_processing/real_time/api.py

@@ -0,0 +1,57 @@
+"""Local only processors for handling real time object processing."""
+
+import logging
+from abc import ABC, abstractmethod
+
+import numpy as np
+
+from frigate.config import FrigateConfig
+
+from ..types import DataProcessorMetrics
+
+logger = logging.getLogger(__name__)
+
+
+class RealTimeProcessorApi(ABC):
+    @abstractmethod
+    def __init__(self, config: FrigateConfig, metrics: DataProcessorMetrics) -> None:
+        self.config = config
+        self.metrics = metrics
+        pass
+
+    @abstractmethod
+    def process_frame(self, obj_data: dict[str, any], frame: np.ndarray) -> None:
+        """Processes the frame with object data.
+        Args:
+            obj_data (dict): containing data about focused object in frame.
+            frame (ndarray): full yuv frame.
+
+        Returns:
+            None.
+        """
+        pass
+
+    @abstractmethod
+    def handle_request(
+        self, topic: str, request_data: dict[str, any]
+    ) -> dict[str, any] | None:
+        """Handle metadata requests.
+        Args:
+            topic (str): topic that dictates what work is requested.
+            request_data (dict): containing data about requested change to process.
+
+        Returns:
+            None if request was not handled, otherwise return response.
+        """
+        pass
+
+    @abstractmethod
+    def expire_object(self, object_id: str) -> None:
+        """Handle objects that are no longer detected.
+        Args:
+            object_id (str): id of object that is no longer detected.
+
+        Returns:
+            None.
+        """
+        pass

frigate/data_processing/real_time/bird_processor.py

@@ -0,0 +1,154 @@
+"""Handle processing images to classify birds."""
+
+import logging
+import os
+
+import cv2
+import numpy as np
+import requests
+
+from frigate.config import FrigateConfig
+from frigate.const import FRIGATE_LOCALHOST, MODEL_CACHE_DIR
+from frigate.util.object import calculate_region
+
+from ..types import DataProcessorMetrics
+from .api import RealTimeProcessorApi
+
+try:
+    from tflite_runtime.interpreter import Interpreter
+except ModuleNotFoundError:
+    from tensorflow.lite.python.interpreter import Interpreter
+
+logger = logging.getLogger(__name__)
+
+
+class BirdProcessor(RealTimeProcessorApi):
+    def __init__(self, config: FrigateConfig, metrics: DataProcessorMetrics):
+        super().__init__(config, metrics)
+        self.interpreter: Interpreter = None
+        self.tensor_input_details: dict[str, any] = None
+        self.tensor_output_details: dict[str, any] = None
+        self.detected_birds: dict[str, float] = {}
+        self.labelmap: dict[int, str] = {}
+
+        download_path = os.path.join(MODEL_CACHE_DIR, "bird")
+        self.model_files = {
+            "bird.tflite": "https://raw.githubusercontent.com/google-coral/test_data/master/mobilenet_v2_1.0_224_inat_bird_quant.tflite",
+            "birdmap.txt": "https://raw.githubusercontent.com/google-coral/test_data/master/inat_bird_labels.txt",
+        }
+
+        if not all(
+            os.path.exists(os.path.join(download_path, n))
+            for n in self.model_files.keys()
+        ):
+            # conditionally import ModelDownloader
+            from frigate.util.downloader import ModelDownloader
+
+            self.downloader = ModelDownloader(
+                model_name="bird",
+                download_path=download_path,
+                file_names=self.model_files.keys(),
+                download_func=self.__download_models,
+                complete_func=self.__build_detector,
+            )
+            self.downloader.ensure_model_files()
+        else:
+            self.__build_detector()
+
+    def __download_models(self, path: str) -> None:
+        try:
+            file_name = os.path.basename(path)
+
+            # conditionally import ModelDownloader
+            from frigate.util.downloader import ModelDownloader
+
+            ModelDownloader.download_from_url(self.model_files[file_name], path)
+        except Exception as e:
+            logger.error(f"Failed to download {path}: {e}")
+
+    def __build_detector(self) -> None:
+        self.interpreter = Interpreter(
+            model_path=os.path.join(MODEL_CACHE_DIR, "bird/bird.tflite"),
+            num_threads=2,
+        )
+        self.interpreter.allocate_tensors()
+        self.tensor_input_details = self.interpreter.get_input_details()
+        self.tensor_output_details = self.interpreter.get_output_details()
+
+        i = 0
+
+        with open(os.path.join(MODEL_CACHE_DIR, "bird/birdmap.txt")) as f:
+            line = f.readline()
+            while line:
+                start = line.find("(")
+                end = line.find(")")
+                self.labelmap[i] = line[start + 1 : end]
+                i += 1
+                line = f.readline()
+
+    def process_frame(self, obj_data, frame):
+        if obj_data["label"] != "bird":
+            return
+
+        x, y, x2, y2 = calculate_region(
+            frame.shape,
+            obj_data["box"][0],
+            obj_data["box"][1],
+            obj_data["box"][2],
+            obj_data["box"][3],
+            224,
+            1.0,
+        )
+
+        rgb = cv2.cvtColor(frame, cv2.COLOR_YUV2RGB_I420)
+        input = rgb[
+            y:y2,
+            x:x2,
+        ]
+
+        cv2.imwrite("/media/frigate/test_class.png", input)
+
+        input = np.expand_dims(input, axis=0)
+
+        self.interpreter.set_tensor(self.tensor_input_details[0]["index"], input)
+        self.interpreter.invoke()
+        res: np.ndarray = self.interpreter.get_tensor(
+            self.tensor_output_details[0]["index"]
+        )[0]
+        probs = res / res.sum(axis=0)
+        best_id = np.argmax(probs)
+
+        if best_id == 964:
+            logger.debug("No bird classification was detected.")
+            return
+
+        score = round(probs[best_id], 2)
+
+        if score < self.config.classification.bird.threshold:
+            logger.debug(f"Score {score} is not above required threshold")
+            return
+
+        previous_score = self.detected_birds.get(obj_data["id"], 0.0)
+
+        if score <= previous_score:
+            logger.debug(f"Score {score} is worse than previous score {previous_score}")
+            return
+
+        resp = requests.post(
+            f"{FRIGATE_LOCALHOST}/api/events/{obj_data['id']}/sub_label",
+            json={
+                "camera": obj_data.get("camera"),
+                "subLabel": self.labelmap[best_id],
+                "subLabelScore": score,
+            },
+        )
+
+        if resp.status_code == 200:
+            self.detected_birds[obj_data["id"]] = score
+
+    def handle_request(self, topic, request_data):
+        return None
+
+    def expire_object(self, object_id):
+        if object_id in self.detected_birds:
+            self.detected_birds.pop(object_id)

frigate/data_processing/real_time/face_processor.py

@@ -0,0 +1,406 @@
+"""Handle processing images for face detection and recognition."""
+
+import base64
+import datetime
+import logging
+import os
+import random
+import string
+from typing import Optional
+
+import cv2
+import numpy as np
+import requests
+
+from frigate.comms.embeddings_updater import EmbeddingsRequestEnum
+from frigate.config import FrigateConfig
+from frigate.const import FACE_DIR, FRIGATE_LOCALHOST, MODEL_CACHE_DIR
+from frigate.util.image import area
+
+from ..types import DataProcessorMetrics
+from .api import RealTimeProcessorApi
+
+logger = logging.getLogger(__name__)
+
+
+MIN_MATCHING_FACES = 2
+
+
+class FaceProcessor(RealTimeProcessorApi):
+    def __init__(self, config: FrigateConfig, metrics: DataProcessorMetrics):
+        super().__init__(config, metrics)
+        self.face_config = config.face_recognition
+        self.face_detector: cv2.FaceDetectorYN = None
+        self.landmark_detector: cv2.face.FacemarkLBF = None
+        self.face_recognizer: cv2.face.LBPHFaceRecognizer = None
+        self.requires_face_detection = "face" not in self.config.objects.all_objects
+        self.detected_faces: dict[str, float] = {}
+
+        download_path = os.path.join(MODEL_CACHE_DIR, "facedet")
+        self.model_files = {
+            "facedet.onnx": "https://github.com/NickM-27/facenet-onnx/releases/download/v1.0/facedet.onnx",
+            "landmarkdet.yaml": "https://github.com/NickM-27/facenet-onnx/releases/download/v1.0/landmarkdet.yaml",
+        }
+
+        if not all(
+            os.path.exists(os.path.join(download_path, n))
+            for n in self.model_files.keys()
+        ):
+            # conditionally import ModelDownloader
+            from frigate.util.downloader import ModelDownloader
+
+            self.downloader = ModelDownloader(
+                model_name="facedet",
+                download_path=download_path,
+                file_names=self.model_files.keys(),
+                download_func=self.__download_models,
+                complete_func=self.__build_detector,
+            )
+            self.downloader.ensure_model_files()
+        else:
+            self.__build_detector()
+
+        self.label_map: dict[int, str] = {}
+        self.__build_classifier()
+
+    def __download_models(self, path: str) -> None:
+        try:
+            file_name = os.path.basename(path)
+            # conditionally import ModelDownloader
+            from frigate.util.downloader import ModelDownloader
+
+            ModelDownloader.download_from_url(self.model_files[file_name], path)
+        except Exception as e:
+            logger.error(f"Failed to download {path}: {e}")
+
+    def __build_detector(self) -> None:
+        self.face_detector = cv2.FaceDetectorYN.create(
+            "/config/model_cache/facedet/facedet.onnx",
+            config="",
+            input_size=(320, 320),
+            score_threshold=0.8,
+            nms_threshold=0.3,
+        )
+        self.landmark_detector = cv2.face.createFacemarkLBF()
+        self.landmark_detector.loadModel("/config/model_cache/facedet/landmarkdet.yaml")
+
+    def __build_classifier(self) -> None:
+        if not self.landmark_detector:
+            return None
+
+        labels = []
+        faces = []
+
+        dir = "/media/frigate/clips/faces"
+        for idx, name in enumerate(os.listdir(dir)):
+            if name == "train":
+                continue
+
+            face_folder = os.path.join(dir, name)
+
+            if not os.path.isdir(face_folder):
+                continue
+
+            self.label_map[idx] = name
+            for image in os.listdir(face_folder):
+                img = cv2.imread(os.path.join(face_folder, image))
+
+                if img is None:
+                    continue
+
+                img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
+                img = self.__align_face(img, img.shape[1], img.shape[0])
+                faces.append(img)
+                labels.append(idx)
+
+        self.recognizer: cv2.face.LBPHFaceRecognizer = (
+            cv2.face.LBPHFaceRecognizer_create(
+                radius=2, threshold=(1 - self.face_config.min_score) * 1000
+            )
+        )
+        self.recognizer.train(faces, np.array(labels))
+
+    def __align_face(
+        self,
+        image: np.ndarray,
+        output_width: int,
+        output_height: int,
+    ) -> np.ndarray:
+        _, lands = self.landmark_detector.fit(
+            image, np.array([(0, 0, image.shape[1], image.shape[0])])
+        )
+        landmarks: np.ndarray = lands[0][0]
+
+        # get landmarks for eyes
+        leftEyePts = landmarks[42:48]
+        rightEyePts = landmarks[36:42]
+
+        # compute the center of mass for each eye
+        leftEyeCenter = leftEyePts.mean(axis=0).astype("int")
+        rightEyeCenter = rightEyePts.mean(axis=0).astype("int")
+
+        # compute the angle between the eye centroids
+        dY = rightEyeCenter[1] - leftEyeCenter[1]
+        dX = rightEyeCenter[0] - leftEyeCenter[0]
+        angle = np.degrees(np.arctan2(dY, dX)) - 180
+
+        # compute the desired right eye x-coordinate based on the
+        # desired x-coordinate of the left eye
+        desiredRightEyeX = 1.0 - 0.35
+
+        # determine the scale of the new resulting image by taking
+        # the ratio of the distance between eyes in the *current*
+        # image to the ratio of distance between eyes in the
+        # *desired* image
+        dist = np.sqrt((dX**2) + (dY**2))
+        desiredDist = desiredRightEyeX - 0.35
+        desiredDist *= output_width
+        scale = desiredDist / dist
+
+        # compute center (x, y)-coordinates (i.e., the median point)
+        # between the two eyes in the input image
+        # grab the rotation matrix for rotating and scaling the face
+        eyesCenter = (
+            int((leftEyeCenter[0] + rightEyeCenter[0]) // 2),
+            int((leftEyeCenter[1] + rightEyeCenter[1]) // 2),
+        )
+        M = cv2.getRotationMatrix2D(eyesCenter, angle, scale)
+
+        # update the translation component of the matrix
+        tX = output_width * 0.5
+        tY = output_height * 0.35
+        M[0, 2] += tX - eyesCenter[0]
+        M[1, 2] += tY - eyesCenter[1]
+
+        # apply the affine transformation
+        return cv2.warpAffine(
+            image, M, (output_width, output_height), flags=cv2.INTER_CUBIC
+        )
+
+    def __clear_classifier(self) -> None:
+        self.face_recognizer = None
+        self.label_map = {}
+
+    def __detect_face(self, input: np.ndarray) -> tuple[int, int, int, int]:
+        """Detect faces in input image."""
+        if not self.face_detector:
+            return None
+
+        self.face_detector.setInputSize((input.shape[1], input.shape[0]))
+        faces = self.face_detector.detect(input)
+
+        if faces is None or faces[1] is None:
+            return None
+
+        face = None
+
+        for _, potential_face in enumerate(faces[1]):
+            raw_bbox = potential_face[0:4].astype(np.uint16)
+            x: int = max(raw_bbox[0], 0)
+            y: int = max(raw_bbox[1], 0)
+            w: int = raw_bbox[2]
+            h: int = raw_bbox[3]
+            bbox = (x, y, x + w, y + h)
+
+            if face is None or area(bbox) > area(face):
+                face = bbox
+
+        return face
+
+    def __classify_face(self, face_image: np.ndarray) -> tuple[str, float] | None:
+        if not self.landmark_detector:
+            return None
+
+        if not self.label_map:
+            self.__build_classifier()
+
+        img = cv2.cvtColor(face_image, cv2.COLOR_BGR2GRAY)
+        img = self.__align_face(img, img.shape[1], img.shape[0])
+        index, distance = self.recognizer.predict(img)
+
+        if index == -1:
+            return None
+
+        score = 1.0 - (distance / 1000)
+        return self.label_map[index], round(score, 2)
+
+    def __update_metrics(self, duration: float) -> None:
+        self.metrics.face_rec_fps.value = (
+            self.metrics.face_rec_fps.value * 9 + duration
+        ) / 10
+
+    def process_frame(self, obj_data: dict[str, any], frame: np.ndarray):
+        """Look for faces in image."""
+        start = datetime.datetime.now().timestamp()
+        id = obj_data["id"]
+
+        # don't run for non person objects
+        if obj_data.get("label") != "person":
+            logger.debug("Not a processing face for non person object.")
+            return
+
+        # don't overwrite sub label for objects that have a sub label
+        # that is not a face
+        if obj_data.get("sub_label") and id not in self.detected_faces:
+            logger.debug(
+                f"Not processing face due to existing sub label: {obj_data.get('sub_label')}."
+            )
+            return
+
+        face: Optional[dict[str, any]] = None
+
+        if self.requires_face_detection:
+            logger.debug("Running manual face detection.")
+            person_box = obj_data.get("box")
+
+            if not person_box:
+                return
+
+            rgb = cv2.cvtColor(frame, cv2.COLOR_YUV2RGB_I420)
+            left, top, right, bottom = person_box
+            person = rgb[top:bottom, left:right]
+            face_box = self.__detect_face(person)
+
+            if not face_box:
+                logger.debug("Detected no faces for person object.")
+                return
+
+            face_frame = person[
+                max(0, face_box[1]) : min(frame.shape[0], face_box[3]),
+                max(0, face_box[0]) : min(frame.shape[1], face_box[2]),
+            ]
+            face_frame = cv2.cvtColor(face_frame, cv2.COLOR_RGB2BGR)
+        else:
+            # don't run for object without attributes
+            if not obj_data.get("current_attributes"):
+                logger.debug("No attributes to parse.")
+                return
+
+            attributes: list[dict[str, any]] = obj_data.get("current_attributes", [])
+            for attr in attributes:
+                if attr.get("label") != "face":
+                    continue
+
+                if face is None or attr.get("score", 0.0) > face.get("score", 0.0):
+                    face = attr
+
+            # no faces detected in this frame
+            if not face:
+                return
+
+            face_box = face.get("box")
+
+            # check that face is valid
+            if not face_box or area(face_box) < self.config.face_recognition.min_area:
+                logger.debug(f"Invalid face box {face}")
+                return
+
+            face_frame = cv2.cvtColor(frame, cv2.COLOR_YUV2BGR_I420)
+
+            face_frame = face_frame[
+                max(0, face_box[1]) : min(frame.shape[0], face_box[3]),
+                max(0, face_box[0]) : min(frame.shape[1], face_box[2]),
+            ]
+
+        res = self.__classify_face(face_frame)
+
+        if not res:
+            return
+
+        sub_label, score = res
+
+        # calculate the overall face score as the probability * area of face
+        # this will help to reduce false positives from small side-angle faces
+        # if a large front-on face image may have scored slightly lower but
+        # is more likely to be accurate due to the larger face area
+        face_score = round(score * face_frame.shape[0] * face_frame.shape[1], 2)
+
+        logger.debug(
+            f"Detected best face for person as: {sub_label} with probability {score} and overall face score {face_score}"
+        )
+
+        if self.config.face_recognition.save_attempts:
+            # write face to library
+            folder = os.path.join(FACE_DIR, "train")
+            file = os.path.join(folder, f"{id}-{sub_label}-{score}-{face_score}.webp")
+            os.makedirs(folder, exist_ok=True)
+            cv2.imwrite(file, face_frame)
+
+        if score < self.config.face_recognition.threshold:
+            logger.debug(
+                f"Recognized face distance {score} is less than threshold {self.config.face_recognition.threshold}"
+            )
+            self.__update_metrics(datetime.datetime.now().timestamp() - start)
+            return
+
+        if id in self.detected_faces and face_score <= self.detected_faces[id]:
+            logger.debug(
+                f"Recognized face distance {score} and overall score {face_score} is less than previous overall face score ({self.detected_faces.get(id)})."
+            )
+            self.__update_metrics(datetime.datetime.now().timestamp() - start)
+            return
+
+        resp = requests.post(
+            f"{FRIGATE_LOCALHOST}/api/events/{id}/sub_label",
+            json={
+                "camera": obj_data.get("camera"),
+                "subLabel": sub_label,
+                "subLabelScore": score,
+            },
+        )
+
+        if resp.status_code == 200:
+            self.detected_faces[id] = face_score
+
+        self.__update_metrics(datetime.datetime.now().timestamp() - start)
+
+    def handle_request(self, topic, request_data) -> dict[str, any] | None:
+        if topic == EmbeddingsRequestEnum.clear_face_classifier.value:
+            self.__clear_classifier()
+        elif topic == EmbeddingsRequestEnum.register_face.value:
+            rand_id = "".join(
+                random.choices(string.ascii_lowercase + string.digits, k=6)
+            )
+            label = request_data["face_name"]
+            id = f"{label}-{rand_id}"
+
+            if request_data.get("cropped"):
+                thumbnail = request_data["image"]
+            else:
+                img = cv2.imdecode(
+                    np.frombuffer(
+                        base64.b64decode(request_data["image"]), dtype=np.uint8
+                    ),
+                    cv2.IMREAD_COLOR,
+                )
+                face_box = self.__detect_face(img)
+
+                if not face_box:
+                    return {
+                        "message": "No face was detected.",
+                        "success": False,
+                    }
+
+                face = img[face_box[1] : face_box[3], face_box[0] : face_box[2]]
+                _, thumbnail = cv2.imencode(
+                    ".webp", face, [int(cv2.IMWRITE_WEBP_QUALITY), 100]
+                )
+
+            # write face to library
+            folder = os.path.join(FACE_DIR, label)
+            file = os.path.join(folder, f"{id}.webp")
+            os.makedirs(folder, exist_ok=True)
+
+            # save face image
+            with open(file, "wb") as output:
+                output.write(thumbnail.tobytes())
+
+            self.__clear_classifier()
+            return {
+                "message": "Successfully registered face.",
+                "success": True,
+            }
+
+    def expire_object(self, object_id: str):
+        if object_id in self.detected_faces:
+            self.detected_faces.pop(object_id)

frigate/data_processing/types.py

@@ -0,0 +1,24 @@
+"""Embeddings types."""
+
+import multiprocessing as mp
+from enum import Enum
+from multiprocessing.sharedctypes import Synchronized
+
+
+class DataProcessorMetrics:
+    image_embeddings_fps: Synchronized
+    text_embeddings_sps: Synchronized
+    face_rec_fps: Synchronized
+    alpr_pps: Synchronized
+
+    def __init__(self):
+        self.image_embeddings_fps = mp.Value("d", 0.01)
+        self.text_embeddings_sps = mp.Value("d", 0.01)
+        self.face_rec_fps = mp.Value("d", 0.01)
+        self.alpr_pps = mp.Value("d", 0.01)
+
+
+class PostProcessDataEnum(str, Enum):
+    recording = "recording"
+    review = "review"
+    tracked_object = "tracked_object"

frigate/detectors/detector_config.py

@@ -194,6 +194,9 @@ class BaseDetectorConfig(BaseModel):
     model: Optional[ModelConfig] = Field(
         default=None, title="Detector specific model configuration."
     )
+    model_path: Optional[str] = Field(
+        default=None, title="Detector specific model path."
+    )
     model_config = ConfigDict(
         extra="allow", arbitrary_types_allowed=True, protected_namespaces=()
     )

frigate/detectors/plugins/rknn.py

@@ -108,7 +108,7 @@ class Rknn(DetectionApi):
                     model_props["model_type"] = model_type
 
             if model_matched:
-                model_props["filename"] = model_path + f"-{soc}-v2.0.0-1.rknn"
+                model_props["filename"] = model_path + f"-{soc}-v2.3.0-1.rknn"
 
                 model_props["path"] = model_cache_dir + model_props["filename"]
 
@@ -129,7 +129,7 @@ class Rknn(DetectionApi):
             os.mkdir(model_cache_dir)
 
         urllib.request.urlretrieve(
-            f"https://github.com/MarcA711/rknn-models/releases/download/v2.0.0/{filename}",
+            f"https://github.com/MarcA711/rknn-models/releases/download/v2.3.0/{filename}",
             model_cache_dir + filename,
         )
 

frigate/detectors/plugins/tensorrt.py

@@ -219,19 +219,19 @@ class TensorRtDetector(DetectionApi):
         ]
 
     def __init__(self, detector_config: TensorRTDetectorConfig):
-        assert (
-            TRT_SUPPORT
-        ), f"TensorRT libraries not found, {DETECTOR_KEY} detector not present"
+        assert TRT_SUPPORT, (
+            f"TensorRT libraries not found, {DETECTOR_KEY} detector not present"
+        )
 
         (cuda_err,) = cuda.cuInit(0)
-        assert (
-            cuda_err == cuda.CUresult.CUDA_SUCCESS
-        ), f"Failed to initialize cuda {cuda_err}"
+        assert cuda_err == cuda.CUresult.CUDA_SUCCESS, (
+            f"Failed to initialize cuda {cuda_err}"
+        )
         err, dev_count = cuda.cuDeviceGetCount()
         logger.debug(f"Num Available Devices: {dev_count}")
-        assert (
-            detector_config.device < dev_count
-        ), f"Invalid TensorRT Device Config. Device {detector_config.device} Invalid."
+        assert detector_config.device < dev_count, (
+            f"Invalid TensorRT Device Config. Device {detector_config.device} Invalid."
+        )
         err, self.cu_ctx = cuda.cuCtxCreate(
             cuda.CUctx_flags.CU_CTX_MAP_HOST, detector_config.device
         )

frigate/embeddings/__init__.py

@@ -1,5 +1,6 @@
 """SQLite-vec embeddings database."""
 
+import base64
 import json
 import logging
 import multiprocessing as mp
@@ -13,7 +14,8 @@ from setproctitle import setproctitle
 
 from frigate.comms.embeddings_updater import EmbeddingsRequestEnum, EmbeddingsRequestor
 from frigate.config import FrigateConfig
-from frigate.const import CONFIG_DIR
+from frigate.const import CONFIG_DIR, FACE_DIR
+from frigate.data_processing.types import DataProcessorMetrics
 from frigate.db.sqlitevecq import SqliteVecQueueDatabase
 from frigate.models import Event
 from frigate.util.builtin import serialize
@@ -25,7 +27,7 @@ from .util import ZScoreNormalization
 logger = logging.getLogger(__name__)
 
 
-def manage_embeddings(config: FrigateConfig) -> None:
+def manage_embeddings(config: FrigateConfig, metrics: DataProcessorMetrics) -> None:
     # Only initialize embeddings if semantic search is enabled
     if not config.semantic_search.enabled:
         return
@@ -59,6 +61,7 @@ def manage_embeddings(config: FrigateConfig) -> None:
     maintainer = EmbeddingMaintainer(
         db,
         config,
+        metrics,
         stop_event,
     )
     maintainer.start()
@@ -189,6 +192,38 @@ class EmbeddingsContext:
 
         return results
 
+    def register_face(self, face_name: str, image_data: bytes) -> dict[str, any]:
+        return self.requestor.send_data(
+            EmbeddingsRequestEnum.register_face.value,
+            {
+                "face_name": face_name,
+                "image": base64.b64encode(image_data).decode("ASCII"),
+            },
+        )
+
+    def get_face_ids(self, name: str) -> list[str]:
+        sql_query = f"""
+            SELECT
+                id
+            FROM vec_descriptions
+            WHERE id LIKE '%{name}%'
+        """
+
+        return self.db.execute_sql(sql_query).fetchall()
+
+    def clear_face_classifier(self) -> None:
+        self.requestor.send_data(
+            EmbeddingsRequestEnum.clear_face_classifier.value, None
+        )
+
+    def delete_face_ids(self, face: str, ids: list[str]) -> None:
+        folder = os.path.join(FACE_DIR, face)
+        for id in ids:
+            file_path = os.path.join(folder, id)
+
+            if os.path.isfile(file_path):
+                os.unlink(file_path)
+
     def update_description(self, event_id: str, description: str) -> None:
         self.requestor.send_data(
             EmbeddingsRequestEnum.embed_description.value,