ensure detection_start doesnt change values between conditions

Dockerfile

@@ -25,7 +25,6 @@ RUN apt -qq update && apt -qq install --no-install-recommends -y \
         imutils \
         scipy \
     && python3.7 -m pip install -U \
-        SharedArray \
         Flask \
         paho-mqtt \
         PyYAML \
@@ -38,9 +37,9 @@ RUN apt -qq update && apt -qq install --no-install-recommends -y \
     && apt -qq install --no-install-recommends -y \
         libedgetpu1-max \
     ## Tensorflow lite (python 3.7 only)
-    && wget -q https://dl.google.com/coral/python/tflite_runtime-2.1.0-cp37-cp37m-linux_x86_64.whl \
+    && wget -q https://dl.google.com/coral/python/tflite_runtime-2.1.0.post1-cp37-cp37m-linux_x86_64.whl \
-    && python3.7 -m pip install tflite_runtime-2.1.0-cp37-cp37m-linux_x86_64.whl \
+    && python3.7 -m pip install tflite_runtime-2.1.0.post1-cp37-cp37m-linux_x86_64.whl \
-    && rm tflite_runtime-2.1.0-cp37-cp37m-linux_x86_64.whl \
+    && rm tflite_runtime-2.1.0.post1-cp37-cp37m-linux_x86_64.whl \
     && rm -rf /var/lib/apt/lists/* \
     && (apt-get autoremove -y; apt-get autoclean -y)
 

README.md

@@ -16,16 +16,6 @@ You see multiple bounding boxes because it draws bounding boxes from all frames
 [![](http://img.youtube.com/vi/nqHbCtyo4dY/0.jpg)](http://www.youtube.com/watch?v=nqHbCtyo4dY "Frigate")
 
 ## Getting Started
-Build the container with
-```
-docker build -t frigate .
-```
-
-Models for both CPU and EdgeTPU (Coral) are bundled in the image. You can use your own models with volume mounts:
-- CPU Model: `/cpu_model.tflite`
-- EdgeTPU Model: `/edgetpu_model.tflite`
-- Labels: `/labelmap.txt`
-
 Run the container with
 ```bash
 docker run --rm \
@@ -36,7 +26,7 @@ docker run --rm \
 -v /etc/localtime:/etc/localtime:ro \
 -p 5000:5000 \
 -e FRIGATE_RTSP_PASSWORD='password' \
-frigate:latest
+blakeblackshear/frigate:stable
 ```
 
 Example docker-compose:
@@ -46,7 +36,7 @@ Example docker-compose:
     restart: unless-stopped
     privileged: true
     shm_size: '1g' # should work for 5-7 cameras
-    image: frigate:latest
+    image: blakeblackshear/frigate:stable
     volumes:
       - /dev/bus/usb:/dev/bus/usb
       - /etc/localtime:/etc/localtime:ro
@@ -127,6 +117,11 @@ sensor:
         value_template: '{{ states.sensor.frigate_debug.attributes["coral"]["inference_speed"] }}' 
         unit_of_measurement: 'ms'
 ```
+## Using a custom model
+Models for both CPU and EdgeTPU (Coral) are bundled in the image. You can use your own models with volume mounts:
+- CPU Model: `/cpu_model.tflite`
+- EdgeTPU Model: `/edgetpu_model.tflite`
+- Labels: `/labelmap.txt`
 
 ## Tips
 - Lower the framerate of the video feed on the camera to reduce the CPU usage for capturing the feed

benchmark.py

@@ -1,18 +1,79 @@
-import statistics
+import os
+from statistics import mean
+import multiprocessing as mp
 import numpy as np
-import time
+import datetime
-from frigate.edgetpu import ObjectDetector
+from frigate.edgetpu import ObjectDetector, EdgeTPUProcess, RemoteObjectDetector, load_labels
 
-object_detector = ObjectDetector()
+my_frame = np.expand_dims(np.full((300,300,3), 1, np.uint8), axis=0)
+labels = load_labels('/labelmap.txt')
 
-frame = np.zeros((300,300,3), np.uint8)
+######
-input_frame = np.expand_dims(frame, axis=0)
+# Minimal same process runner
+######
+# object_detector = ObjectDetector()
+# tensor_input = np.expand_dims(np.full((300,300,3), 0, np.uint8), axis=0)
 
-detection_times = []
+# start = datetime.datetime.now().timestamp()
 
-for x in range(0, 100):
+# frame_times = []
-    start = time.monotonic()
+# for x in range(0, 1000):
-    object_detector.detect_raw(input_frame)
+#   start_frame = datetime.datetime.now().timestamp()
-    detection_times.append(time.monotonic()-start)
 
-print(f"Average inference time: {statistics.mean(detection_times)*1000:.2f}ms")
+#   tensor_input[:] = my_frame
+#   detections = object_detector.detect_raw(tensor_input)
+#   parsed_detections = []
+#   for d in detections:
+#       if d[1] < 0.4:
+#           break
+#       parsed_detections.append((
+#           labels[int(d[0])],
+#           float(d[1]),
+#           (d[2], d[3], d[4], d[5])
+#       ))
+#   frame_times.append(datetime.datetime.now().timestamp()-start_frame)
+
+# duration = datetime.datetime.now().timestamp()-start
+# print(f"Processed for {duration:.2f} seconds.")
+# print(f"Average frame processing time: {mean(frame_times)*1000:.2f}ms")
+
+######
+# Separate process runner
+######
+def start(id, num_detections, detection_queue):
+  object_detector = RemoteObjectDetector(str(id), '/labelmap.txt', detection_queue)
+  start = datetime.datetime.now().timestamp()
+
+  frame_times = []
+  for x in range(0, num_detections):
+    start_frame = datetime.datetime.now().timestamp()
+    detections = object_detector.detect(my_frame)
+    frame_times.append(datetime.datetime.now().timestamp()-start_frame)
+
+  duration = datetime.datetime.now().timestamp()-start
+  print(f"{id} - Processed for {duration:.2f} seconds.")
+  print(f"{id} - Average frame processing time: {mean(frame_times)*1000:.2f}ms")
+
+edgetpu_process = EdgeTPUProcess()
+
+# start(1, 1000, edgetpu_process.detect_lock, edgetpu_process.detect_ready, edgetpu_process.frame_ready)
+
+####
+# Multiple camera processes
+####
+camera_processes = []
+for x in range(0, 10):
+  camera_process = mp.Process(target=start, args=(x, 100, edgetpu_process.detection_queue))
+  camera_process.daemon = True
+  camera_processes.append(camera_process)
+
+start = datetime.datetime.now().timestamp()
+
+for p in camera_processes:
+  p.start()
+
+for p in camera_processes:
+  p.join()
+
+duration = datetime.datetime.now().timestamp()-start
+print(f"Total - Processed for {duration:.2f} seconds.")

config/config.example.yml

@@ -3,9 +3,13 @@ web_port: 5000
 mqtt:
   host: mqtt.server.com
   topic_prefix: frigate
-#  client_id: frigate # Optional -- set to override default client id of 'frigate' if running multiple instances
+  # client_id: frigate # Optional -- set to override default client id of 'frigate' if running multiple instances
-#  user: username # Optional -- Uncomment for use
+  # user: username # Optional
-#  password: password # Optional -- Uncomment for use
+  #################
+  ## Environment variables that begin with 'FRIGATE_' may be referenced in {}.
+  ##   password: '{FRIGATE_MQTT_PASSWORD}'
+  #################
+  # password: password # Optional
 
 #################
 # Default ffmpeg args. Optional and can be overwritten per camera.
@@ -106,13 +110,6 @@ cameras:
     ################
     take_frame: 1
 
-    ################
-    # The expected framerate for the camera. Frigate will try and ensure it maintains this framerate
-    # by dropping frames as necessary. Setting this lower than the actual framerate will allow frigate
-    # to process every frame at the expense of realtime processing.
-    ################
-    fps: 5
-
     ################
     # Configuration for the snapshots in the debug view and mqtt
     ################

detect_objects.py

@@ -1,3 +1,7 @@
+import os
+import sys
+import traceback
+import signal
 import cv2
 import time
 import datetime
@@ -8,14 +12,16 @@ import multiprocessing as mp
 import subprocess as sp
 import numpy as np
 import logging
-from flask import Flask, Response, make_response, jsonify
+from flask import Flask, Response, make_response, jsonify, request
 import paho.mqtt.client as mqtt
 
-from frigate.video import track_camera
+from frigate.video import track_camera, get_ffmpeg_input, get_frame_shape, CameraCapture, start_or_restart_ffmpeg
 from frigate.object_processing import TrackedObjectProcessor
 from frigate.util import EventsPerSecond
 from frigate.edgetpu import EdgeTPUProcess
 
+FRIGATE_VARS = {k: v for k, v in os.environ.items() if k.startswith('FRIGATE_')}
+
 with open('/config/config.yml') as f:
     CONFIG = yaml.safe_load(f)
 
@@ -24,6 +30,8 @@ MQTT_PORT = CONFIG.get('mqtt', {}).get('port', 1883)
 MQTT_TOPIC_PREFIX = CONFIG.get('mqtt', {}).get('topic_prefix', 'frigate')
 MQTT_USER = CONFIG.get('mqtt', {}).get('user')
 MQTT_PASS = CONFIG.get('mqtt', {}).get('password')
+if not MQTT_PASS is None:
+    MQTT_PASS = MQTT_PASS.format(**FRIGATE_VARS)
 MQTT_CLIENT_ID = CONFIG.get('mqtt', {}).get('client_id', 'frigate')
 
 # Set the default FFmpeg config
@@ -31,7 +39,7 @@ FFMPEG_CONFIG = CONFIG.get('ffmpeg', {})
 FFMPEG_DEFAULT_CONFIG = {
     'global_args': FFMPEG_CONFIG.get('global_args', 
         ['-hide_banner','-loglevel','panic']),
-    'hwaccel_args': FFMPEG_CONFIG.get('hwaccel_args', 
+    'hwaccel_args': FFMPEG_CONFIG.get('hwaccel_args',
         []),
     'input_args': FFMPEG_CONFIG.get('input_args',
         ['-avoid_negative_ts', 'make_zero',
@@ -53,44 +61,72 @@ GLOBAL_OBJECT_CONFIG = CONFIG.get('objects', {})
 WEB_PORT = CONFIG.get('web_port', 5000)
 DEBUG = (CONFIG.get('debug', '0') == '1')
 
+def start_plasma_store():
+    plasma_cmd = ['plasma_store', '-m', '400000000', '-s', '/tmp/plasma']
+    plasma_process = sp.Popen(plasma_cmd, stdout=sp.DEVNULL, stderr=sp.DEVNULL)
+    time.sleep(1)
+    rc = plasma_process.poll()
+    if rc is not None:
+        return None
+    return plasma_process
+
 class CameraWatchdog(threading.Thread):
-    def __init__(self, camera_processes, config, tflite_process, tracked_objects_queue, object_processor):
+    def __init__(self, camera_processes, config, tflite_process, tracked_objects_queue, plasma_process):
         threading.Thread.__init__(self)
         self.camera_processes = camera_processes
         self.config = config
         self.tflite_process = tflite_process
         self.tracked_objects_queue = tracked_objects_queue
-        self.object_processor = object_processor
+        self.plasma_process = plasma_process
 
     def run(self):
         time.sleep(10)
         while True:
             # wait a bit before checking
             time.sleep(10)
+            
+            # check the plasma process
+            rc = self.plasma_process.poll()
+            if rc != None:
+                print(f"plasma_process exited unexpectedly with {rc}")
+                self.plasma_process = start_plasma_store()
 
+            # check the detection process
+            detection_start = self.tflite_process.detection_start.value
+            if (detection_start > 0.0 and 
+                datetime.datetime.now().timestamp() - detection_start > 10):
+                print("Detection appears to be stuck. Restarting detection process")
+                self.tflite_process.start_or_restart()
+            elif not self.tflite_process.detect_process.is_alive():
+                print("Detection appears to have stopped. Restarting detection process")
+                self.tflite_process.start_or_restart()
+
+            # check the camera processes
             for name, camera_process in self.camera_processes.items():
                 process = camera_process['process']
-                if (datetime.datetime.now().timestamp() - self.object_processor.get_current_frame_time(name)) > 30:
-                    print(f"Last frame for {name} is more than 30 seconds old...")
-                    if process.is_alive():
-                        process.terminate()
-                        print("Waiting for process to exit gracefully...")
-                        process.join(timeout=30)
-                        if process.exitcode is None:
-                            print("Process didnt exit. Force killing...")
-                            process.kill()
-                            process.join()
                 if not process.is_alive():
-                    print(f"Process for {name} is not alive. Starting again...")
+                    print(f"Track process for {name} is not alive. Starting again...")
-                    camera_process['fps'].value = float(self.config[name]['fps'])
+                    camera_process['process_fps'].value = 0.0
-                    camera_process['skipped_fps'].value = 0.0
+                    camera_process['detection_fps'].value = 0.0
-                    process = mp.Process(target=track_camera, args=(name, self.config[name], FFMPEG_DEFAULT_CONFIG, GLOBAL_OBJECT_CONFIG, 
+                    camera_process['read_start'].value = 0.0
-                        self.tflite_process.detect_lock, self.tflite_process.detect_ready, self.tflite_process.frame_ready, self.tracked_objects_queue, 
+                    process = mp.Process(target=track_camera, args=(name, self.config[name], GLOBAL_OBJECT_CONFIG, camera_process['frame_queue'],
-                        camera_process['fps'], camera_process['skipped_fps'], camera_process['detection_fps']))
+                        camera_process['frame_shape'], self.tflite_process.detection_queue, self.tracked_objects_queue, 
+                        camera_process['process_fps'], camera_process['detection_fps'],
+                        camera_process['read_start'], camera_process['detection_frame']))
                     process.daemon = True
                     camera_process['process'] = process
                     process.start()
-                    print(f"Camera_process started for {name}: {process.pid}")
+                    print(f"Track process started for {name}: {process.pid}")
+                
+                if not camera_process['capture_thread'].is_alive():
+                    frame_shape = camera_process['frame_shape']
+                    frame_size = frame_shape[0] * frame_shape[1] * frame_shape[2]
+                    ffmpeg_process = start_or_restart_ffmpeg(camera_process['ffmpeg_cmd'], frame_size)
+                    camera_capture = CameraCapture(name, ffmpeg_process, frame_shape, camera_process['frame_queue'], 
+                        camera_process['take_frame'], camera_process['camera_fps'], camera_process['detection_frame'])
+                    camera_capture.start()
+                    camera_process['ffmpeg_process'] = ffmpeg_process
+                    camera_process['capture_thread'] = camera_capture
 
 def main():
     # connect to mqtt and setup last will
@@ -115,14 +151,7 @@ def main():
     client.connect(MQTT_HOST, MQTT_PORT, 60)
     client.loop_start()
 
-    # start plasma store
+    plasma_process = start_plasma_store()
-    plasma_cmd = ['plasma_store', '-m', '400000000', '-s', '/tmp/plasma']
-    plasma_process = sp.Popen(plasma_cmd, stdout=sp.DEVNULL, stderr=sp.DEVNULL)
-    time.sleep(1)
-    rc = plasma_process.poll()
-    if rc is not None:
-        raise RuntimeError("plasma_store exited unexpectedly with "
-                            "code %d" % (rc,))
 
     ##
     # Setup config defaults for cameras
@@ -133,7 +162,7 @@ def main():
         }
 
     # Queue for cameras to push tracked objects to
-    tracked_objects_queue = mp.Queue()
+    tracked_objects_queue = mp.SimpleQueue()
     
     # Start the shared tflite process
     tflite_process = EdgeTPUProcess()
@@ -141,15 +170,56 @@ def main():
     # start the camera processes
     camera_processes = {}
     for name, config in CONFIG['cameras'].items():
+        # Merge the ffmpeg config with the global config
+        ffmpeg = config.get('ffmpeg', {})
+        ffmpeg_input = get_ffmpeg_input(ffmpeg['input'])
+        ffmpeg_global_args = ffmpeg.get('global_args', FFMPEG_DEFAULT_CONFIG['global_args'])
+        ffmpeg_hwaccel_args = ffmpeg.get('hwaccel_args', FFMPEG_DEFAULT_CONFIG['hwaccel_args'])
+        ffmpeg_input_args = ffmpeg.get('input_args', FFMPEG_DEFAULT_CONFIG['input_args'])
+        ffmpeg_output_args = ffmpeg.get('output_args', FFMPEG_DEFAULT_CONFIG['output_args'])
+        ffmpeg_cmd = (['ffmpeg'] +
+                ffmpeg_global_args +
+                ffmpeg_hwaccel_args +
+                ffmpeg_input_args +
+                ['-i', ffmpeg_input] +
+                ffmpeg_output_args +
+                ['pipe:'])
+        
+        if 'width' in config and 'height' in config:
+            frame_shape = (config['height'], config['width'], 3)
+        else:
+            frame_shape = get_frame_shape(ffmpeg_input)
+
+        frame_size = frame_shape[0] * frame_shape[1] * frame_shape[2]
+        take_frame = config.get('take_frame', 1)
+
+        detection_frame = mp.Value('d', 0.0)
+
+        ffmpeg_process = start_or_restart_ffmpeg(ffmpeg_cmd, frame_size)
+        frame_queue = mp.SimpleQueue()
+        camera_fps = EventsPerSecond()
+        camera_fps.start()
+        camera_capture = CameraCapture(name, ffmpeg_process, frame_shape, frame_queue, take_frame, camera_fps, detection_frame)
+        camera_capture.start()
+
         camera_processes[name] = {
-            'fps': mp.Value('d', float(config['fps'])),
+            'camera_fps': camera_fps,
-            'skipped_fps': mp.Value('d', 0.0),
+            'take_frame': take_frame,
+            'process_fps': mp.Value('d', 0.0),
             'detection_fps': mp.Value('d', 0.0),
-            'last_frame': datetime.datetime.now().timestamp()
+            'detection_frame': detection_frame,
+            'read_start': mp.Value('d', 0.0),
+            'ffmpeg_process': ffmpeg_process,
+            'ffmpeg_cmd': ffmpeg_cmd,
+            'frame_queue': frame_queue,
+            'frame_shape': frame_shape,
+            'capture_thread': camera_capture
         }
-        camera_process = mp.Process(target=track_camera, args=(name, config, FFMPEG_DEFAULT_CONFIG, GLOBAL_OBJECT_CONFIG, 
+
-            tflite_process.detect_lock, tflite_process.detect_ready, tflite_process.frame_ready, tracked_objects_queue, 
+        camera_process = mp.Process(target=track_camera, args=(name, config, GLOBAL_OBJECT_CONFIG, frame_queue, frame_shape,
-            camera_processes[name]['fps'], camera_processes[name]['skipped_fps'], camera_processes[name]['detection_fps']))
+            tflite_process.detection_queue, tracked_objects_queue, camera_processes[name]['process_fps'], 
+            camera_processes[name]['detection_fps'], 
+            camera_processes[name]['read_start'], camera_processes[name]['detection_frame']))
         camera_process.daemon = True
         camera_processes[name]['process'] = camera_process
 
@@ -160,7 +230,7 @@ def main():
     object_processor = TrackedObjectProcessor(CONFIG['cameras'], client, MQTT_TOPIC_PREFIX, tracked_objects_queue)
     object_processor.start()
     
-    camera_watchdog = CameraWatchdog(camera_processes, CONFIG['cameras'], tflite_process, tracked_objects_queue, object_processor)
+    camera_watchdog = CameraWatchdog(camera_processes, CONFIG['cameras'], tflite_process, tracked_objects_queue, plasma_process)
     camera_watchdog.start()
 
     # create a flask app that encodes frames a mjpeg on demand
@@ -173,6 +243,23 @@ def main():
         # return a healh
         return "Frigate is running. Alive and healthy!"
 
+    @app.route('/debug/stack')
+    def processor_stack():
+        frame = sys._current_frames().get(object_processor.ident, None)
+        if frame:
+            return "<br>".join(traceback.format_stack(frame)), 200
+        else:
+            return "no frame found", 200
+
+    @app.route('/debug/print_stack')
+    def print_stack():
+        pid = int(request.args.get('pid', 0))
+        if pid == 0:
+            return "missing pid", 200
+        else:
+            os.kill(pid, signal.SIGUSR1)
+            return "check logs", 200
+
     @app.route('/debug/stats')
     def stats():
         stats = {}
@@ -181,17 +268,32 @@ def main():
 
         for name, camera_stats in camera_processes.items():
             total_detection_fps += camera_stats['detection_fps'].value
+            capture_thread = camera_stats['capture_thread']
             stats[name] = {
-                'fps': camera_stats['fps'].value,
+                'camera_fps': round(capture_thread.fps.eps(), 2),
-                'skipped_fps': camera_stats['skipped_fps'].value,
+                'process_fps': round(camera_stats['process_fps'].value, 2),
-                'detection_fps': camera_stats['detection_fps'].value
+                'skipped_fps': round(capture_thread.skipped_fps.eps(), 2),
+                'detection_fps': round(camera_stats['detection_fps'].value, 2),
+                'read_start': camera_stats['read_start'].value,
+                'pid': camera_stats['process'].pid,
+                'ffmpeg_pid': camera_stats['ffmpeg_process'].pid,
+                'frame_info': {
+                    'read': capture_thread.current_frame,
+                    'detect': camera_stats['detection_frame'].value,
+                    'process': object_processor.camera_data[name]['current_frame_time']
+                }
             }
         
         stats['coral'] = {
-            'fps': total_detection_fps,
+            'fps': round(total_detection_fps, 2),
-            'inference_speed': round(tflite_process.avg_inference_speed.value*1000, 2)
+            'inference_speed': round(tflite_process.avg_inference_speed.value*1000, 2),
+            'detection_start': tflite_process.detection_start.value,
+            'pid': tflite_process.detect_process.pid
         }
 
+        rc = camera_watchdog.plasma_process.poll()
+        stats['plasma_store_rc'] = rc
+
         return jsonify(stats)
 
     @app.route('/<camera_name>/<label>/best.jpg')
@@ -210,28 +312,35 @@ def main():
 
     @app.route('/<camera_name>')
     def mjpeg_feed(camera_name):
+        fps = int(request.args.get('fps', '3'))
+        height = int(request.args.get('h', '360'))
         if camera_name in CONFIG['cameras']:
             # return a multipart response
-            return Response(imagestream(camera_name),
+            return Response(imagestream(camera_name, fps, height),
                             mimetype='multipart/x-mixed-replace; boundary=frame')
         else:
             return "Camera named {} not found".format(camera_name), 404
 
-    def imagestream(camera_name):
+    def imagestream(camera_name, fps, height):
         while True:
-            # max out at 1 FPS
+            # max out at specified FPS
-            time.sleep(1)
+            time.sleep(1/fps)
             frame = object_processor.get_current_frame(camera_name)
             if frame is None:
-                frame = np.zeros((720,1280,3), np.uint8)
+                frame = np.zeros((height,int(height*16/9),3), np.uint8)
+
+            width = int(height*frame.shape[1]/frame.shape[0])
+
+            frame = cv2.resize(frame, dsize=(width, height), interpolation=cv2.INTER_LINEAR)
             frame = cv2.cvtColor(frame, cv2.COLOR_RGB2BGR)
+
             ret, jpg = cv2.imencode('.jpg', frame)
             yield (b'--frame\r\n'
                 b'Content-Type: image/jpeg\r\n\r\n' + jpg.tobytes() + b'\r\n\r\n')
 
     app.run(host='0.0.0.0', port=WEB_PORT, debug=False)
 
-    camera_watchdog.join()
+    object_processor.join()
     
     plasma_process.terminate()
 

frigate/edgetpu.py

@@ -1,11 +1,12 @@
 import os
 import datetime
+import hashlib
 import multiprocessing as mp
 import numpy as np
-import SharedArray as sa
+import pyarrow.plasma as plasma
 import tflite_runtime.interpreter as tflite
 from tflite_runtime.interpreter import load_delegate
-from frigate.util import EventsPerSecond
+from frigate.util import EventsPerSecond, listen
 
 def load_labels(path, encoding='utf-8'):
   """Loads labels from file (with or without index numbers).
@@ -60,77 +61,82 @@ class ObjectDetector():
         
         return detections
 
+def run_detector(detection_queue, avg_speed, start):
+    print(f"Starting detection process: {os.getpid()}")
+    listen()
+    plasma_client = plasma.connect("/tmp/plasma")
+    object_detector = ObjectDetector()
+
+    while True:
+        object_id_str = detection_queue.get()
+        object_id_hash = hashlib.sha1(str.encode(object_id_str))
+        object_id = plasma.ObjectID(object_id_hash.digest())
+        object_id_out = plasma.ObjectID(hashlib.sha1(str.encode(f"out-{object_id_str}")).digest())
+        input_frame = plasma_client.get(object_id, timeout_ms=0)
+
+        if input_frame is plasma.ObjectNotAvailable:
+            continue
+
+        # detect and put the output in the plasma store
+        start.value = datetime.datetime.now().timestamp()
+        plasma_client.put(object_detector.detect_raw(input_frame), object_id_out)
+        duration = datetime.datetime.now().timestamp()-start.value
+        start.value = 0.0
+
+        avg_speed.value = (avg_speed.value*9 + duration)/10
+        
 class EdgeTPUProcess():
     def __init__(self):
-        # TODO: see if we can use the plasma store with a queue and maintain the same speeds
+        self.detection_queue = mp.SimpleQueue()
-        try:
-            sa.delete("frame")
-        except:
-            pass
-        try:
-            sa.delete("detections")
-        except:
-            pass
-
-        self.input_frame = sa.create("frame", shape=(1,300,300,3), dtype=np.uint8)
-        self.detections = sa.create("detections", shape=(20,6), dtype=np.float32)
-
-        self.detect_lock = mp.Lock()
-        self.detect_ready = mp.Event()
-        self.frame_ready = mp.Event()
         self.avg_inference_speed = mp.Value('d', 0.01)
+        self.detection_start = mp.Value('d', 0.0)
+        self.detect_process = None
+        self.start_or_restart()
 
-        def run_detector(detect_ready, frame_ready, avg_speed):
+    def start_or_restart(self):
-            print(f"Starting detection process: {os.getpid()}")
+        self.detection_start.value = 0.0
-            object_detector = ObjectDetector()
+        if (not self.detect_process is None) and self.detect_process.is_alive():
-            input_frame = sa.attach("frame")
+            self.detect_process.terminate()
-            detections = sa.attach("detections")
+            print("Waiting for detection process to exit gracefully...")
-
+            self.detect_process.join(timeout=30)
-            while True:
+            if self.detect_process.exitcode is None:
-                # wait until a frame is ready
+                print("Detection process didnt exit. Force killing...")
-                frame_ready.wait()
+                self.detect_process.kill()
-                start = datetime.datetime.now().timestamp()
+                self.detect_process.join()
-                # signal that the process is busy
+        self.detect_process = mp.Process(target=run_detector, args=(self.detection_queue, self.avg_inference_speed, self.detection_start))
-                frame_ready.clear()
-                detections[:] = object_detector.detect_raw(input_frame)
-                # signal that the process is ready to detect
-                detect_ready.set()
-                duration = datetime.datetime.now().timestamp()-start
-                avg_speed.value = (avg_speed.value*9 + duration)/10
-
-        self.detect_process = mp.Process(target=run_detector, args=(self.detect_ready, self.frame_ready, self.avg_inference_speed))
         self.detect_process.daemon = True
         self.detect_process.start()
 
 class RemoteObjectDetector():
-    def __init__(self, labels, detect_lock, detect_ready, frame_ready):
+    def __init__(self, name, labels, detection_queue):
         self.labels = load_labels(labels)
-
+        self.name = name
-        self.input_frame = sa.attach("frame")
-        self.detections = sa.attach("detections")
-
         self.fps = EventsPerSecond()
-
+        self.plasma_client = plasma.connect("/tmp/plasma")
-        self.detect_lock = detect_lock
+        self.detection_queue = detection_queue
-        self.detect_ready = detect_ready
-        self.frame_ready = frame_ready
     
     def detect(self, tensor_input, threshold=.4):
         detections = []
-        with self.detect_lock:
+
-            self.input_frame[:] = tensor_input
+        now = f"{self.name}-{str(datetime.datetime.now().timestamp())}"
-            # unset detections and signal that a frame is ready
+        object_id_frame = plasma.ObjectID(hashlib.sha1(str.encode(now)).digest())
-            self.detect_ready.clear()
+        object_id_detections = plasma.ObjectID(hashlib.sha1(str.encode(f"out-{now}")).digest())
-            self.frame_ready.set()
+        self.plasma_client.put(tensor_input, object_id_frame)
-            # wait until the detection process is finished,
+        self.detection_queue.put(now)
-            self.detect_ready.wait()
+        raw_detections = self.plasma_client.get(object_id_detections, timeout_ms=10000)
-            for d in self.detections:
+
-                if d[1] < threshold:
+        if raw_detections is plasma.ObjectNotAvailable:
-                    break
+            self.plasma_client.delete([object_id_frame])
-                detections.append((
+            return detections
-                    self.labels[int(d[0])],
+
-                    float(d[1]),
+        for d in raw_detections:
-                    (d[2], d[3], d[4], d[5])
+            if d[1] < threshold:
-                ))
+                break
+            detections.append((
+                self.labels[int(d[0])],
+                float(d[1]),
+                (d[2], d[3], d[4], d[5])
+            ))
+        self.plasma_client.delete([object_id_frame, object_id_detections])
         self.fps.update()
         return detections

frigate/object_processing.py

@@ -1,6 +1,7 @@
 import json
 import hashlib
 import datetime
+import time
 import copy
 import cv2
 import threading
@@ -8,9 +9,8 @@ import numpy as np
 from collections import Counter, defaultdict
 import itertools
 import pyarrow.plasma as plasma
-import SharedArray as sa
 import matplotlib.pyplot as plt
-from frigate.util import draw_box_with_label
+from frigate.util import draw_box_with_label, PlasmaManager
 from frigate.edgetpu import load_labels
 
 PATH_TO_LABELS = '/labelmap.txt'
@@ -29,13 +29,15 @@ class TrackedObjectProcessor(threading.Thread):
         self.client = client
         self.topic_prefix = topic_prefix
         self.tracked_objects_queue = tracked_objects_queue
-        self.plasma_client = plasma.connect("/tmp/plasma")
         self.camera_data = defaultdict(lambda: {
             'best_objects': {},
             'object_status': defaultdict(lambda: defaultdict(lambda: 'OFF')),
             'tracked_objects': {},
-            'current_frame_time': datetime.datetime.now().timestamp()
+            'current_frame': np.zeros((720,1280,3), np.uint8),
+            'current_frame_time': 0.0,
+            'object_id': None
         })
+        self.plasma_client = PlasmaManager()
         
     def get_best(self, camera, label):
         if label in self.camera_data[camera]['best_objects']:
@@ -45,9 +47,6 @@ class TrackedObjectProcessor(threading.Thread):
     
     def get_current_frame(self, camera):
         return self.camera_data[camera]['current_frame']
-    
-    def get_current_frame_time(self, camera):
-        return self.camera_data[camera]['current_frame_time']
 
     def run(self):
         while True:
@@ -57,40 +56,43 @@ class TrackedObjectProcessor(threading.Thread):
             best_objects = self.camera_data[camera]['best_objects']
             current_object_status = self.camera_data[camera]['object_status']
             self.camera_data[camera]['tracked_objects'] = tracked_objects
+            self.camera_data[camera]['current_frame_time'] = frame_time
 
             ###
             # Draw tracked objects on the frame
             ###
-            object_id_hash = hashlib.sha1(str.encode(f"{camera}{frame_time}"))
+            current_frame = self.plasma_client.get(f"{camera}{frame_time}")
-            object_id_bytes = object_id_hash.digest()
-            object_id = plasma.ObjectID(object_id_bytes)
-            current_frame = self.plasma_client.get(object_id)
-
-            # draw the bounding boxes on the frame
-            for obj in tracked_objects.values():
-                thickness = 2
-                color = COLOR_MAP[obj['label']]
-                
-                if obj['frame_time'] != frame_time:
-                    thickness = 1
-                    color = (255,0,0)
 
+            if not current_frame is plasma.ObjectNotAvailable:
                 # draw the bounding boxes on the frame
-                box = obj['box']
+                for obj in tracked_objects.values():
-                draw_box_with_label(current_frame, box[0], box[1], box[2], box[3], obj['label'], f"{int(obj['score']*100)}% {int(obj['area'])}", thickness=thickness, color=color)
+                    thickness = 2
-                # draw the regions on the frame
+                    color = COLOR_MAP[obj['label']]
-                region = obj['region']
+                    
-                cv2.rectangle(current_frame, (region[0], region[1]), (region[2], region[3]), (0,255,0), 1)
+                    if obj['frame_time'] != frame_time:
-            
+                        thickness = 1
-            if config['snapshots']['show_timestamp']:
+                        color = (255,0,0)
-                time_to_show = datetime.datetime.fromtimestamp(frame_time).strftime("%m/%d/%Y %H:%M:%S")
-                cv2.putText(current_frame, time_to_show, (10, 30), cv2.FONT_HERSHEY_SIMPLEX, fontScale=.8, color=(255, 255, 255), thickness=2)
 
-            ###
+                    # draw the bounding boxes on the frame
-            # Set the current frame as ready
+                    box = obj['box']
-            ###
+                    draw_box_with_label(current_frame, box[0], box[1], box[2], box[3], obj['label'], f"{int(obj['score']*100)}% {int(obj['area'])}", thickness=thickness, color=color)
-            self.camera_data[camera]['current_frame'] = current_frame
+                    # draw the regions on the frame
-            self.camera_data[camera]['current_frame_time'] = frame_time
+                    region = obj['region']
+                    cv2.rectangle(current_frame, (region[0], region[1]), (region[2], region[3]), (0,255,0), 1)
+                
+                if config['snapshots']['show_timestamp']:
+                    time_to_show = datetime.datetime.fromtimestamp(frame_time).strftime("%m/%d/%Y %H:%M:%S")
+                    cv2.putText(current_frame, time_to_show, (10, 30), cv2.FONT_HERSHEY_SIMPLEX, fontScale=.8, color=(255, 255, 255), thickness=2)
+
+                ###
+                # Set the current frame
+                ###
+                self.camera_data[camera]['current_frame'] = current_frame
+
+                # delete the previous frame from the plasma store and update the object id
+                if not self.camera_data[camera]['object_id'] is None:
+                    self.plasma_client.delete(self.camera_data[camera]['object_id'])
+                self.camera_data[camera]['object_id'] = f"{camera}{frame_time}"
             
             ###
             # Maintain the highest scoring recent object and frame for each label
@@ -104,10 +106,10 @@ class TrackedObjectProcessor(threading.Thread):
                     # if the object is a higher score than the current best score 
                     # or the current object is more than 1 minute old, use the new object
                     if obj['score'] > best_objects[obj['label']]['score'] or (now - best_objects[obj['label']]['frame_time']) > 60:
-                        obj['frame'] = np.copy(current_frame)
+                        obj['frame'] = np.copy(self.camera_data[camera]['current_frame'])
                         best_objects[obj['label']] = obj
                 else:
-                    obj['frame'] = np.copy(current_frame)
+                    obj['frame'] = np.copy(self.camera_data[camera]['current_frame'])
                     best_objects[obj['label']] = obj
 
             ###
@@ -142,4 +144,4 @@ class TrackedObjectProcessor(threading.Thread):
                 ret, jpg = cv2.imencode('.jpg', best_frame)
                 if ret:
                     jpg_bytes = jpg.tobytes()
-                    self.client.publish(f"{self.topic_prefix}/{camera}/{obj_name}/snapshot", jpg_bytes, retain=True)
+                    self.client.publish(f"{self.topic_prefix}/{camera}/{obj_name}/snapshot", jpg_bytes, retain=True)

frigate/objects.py

@@ -49,14 +49,6 @@ class ObjectTracker():
             obj['history'] = [entry]
 
     def match_and_update(self, frame_time, new_objects):
-        if len(new_objects) == 0:
-            for id in list(self.tracked_objects.keys()):
-                if self.disappeared[id] >= self.max_disappeared:
-                    self.deregister(id)
-                else:
-                    self.disappeared[id] += 1
-            return
-            
         # group by name
         new_object_groups = defaultdict(lambda: [])
         for obj in new_objects:
@@ -69,6 +61,18 @@ class ObjectTracker():
                 'frame_time': frame_time
             })
         
+        # update any tracked objects with labels that are not
+        # seen in the current objects and deregister if needed
+        for obj in list(self.tracked_objects.values()):
+            if not obj['label'] in new_object_groups:
+                if self.disappeared[obj['id']] >= self.max_disappeared:
+                    self.deregister(obj['id'])
+                else:
+                    self.disappeared[obj['id']] += 1
+        
+        if len(new_objects) == 0:
+            return
+        
         # track objects for each label type
         for label, group in new_object_groups.items():
             current_objects = [o for o in self.tracked_objects.values() if o['label'] == label]

frigate/util.py

@@ -1,9 +1,14 @@
 import datetime
+import time
+import signal
+import traceback
 import collections
 import numpy as np
 import cv2
 import threading
 import matplotlib.pyplot as plt
+import hashlib
+import pyarrow.plasma as plasma
 
 def draw_box_with_label(frame, x_min, y_min, x_max, y_max, label, info, thickness=2, color=None, position='ul'):
     if color is None:
@@ -127,3 +132,52 @@ class EventsPerSecond:
         now = datetime.datetime.now().timestamp()
         seconds = min(now-self._start, last_n_seconds)
         return len([t for t in self._timestamps if t > (now-last_n_seconds)]) / seconds
+
+def print_stack(sig, frame):
+    traceback.print_stack(frame)
+
+def listen():
+    signal.signal(signal.SIGUSR1, print_stack)
+
+class PlasmaManager:
+    def __init__(self):
+        self.connect()
+    
+    def connect(self):
+        while True:
+            try:
+                self.plasma_client = plasma.connect("/tmp/plasma")
+                return
+            except:
+                print(f"TrackedObjectProcessor: unable to connect plasma client")
+                time.sleep(10)
+
+    def get(self, name, timeout_ms=0):
+        object_id = plasma.ObjectID(hashlib.sha1(str.encode(name)).digest())
+        while True:
+            try:
+                return self.plasma_client.get(object_id, timeout_ms=timeout_ms)
+            except:
+                self.connect()
+                time.sleep(1)
+
+    def put(self, name, obj):
+        object_id = plasma.ObjectID(hashlib.sha1(str.encode(name)).digest())
+        while True:
+            try:
+                self.plasma_client.put(obj, object_id)
+                return
+            except Exception as e:
+                print(f"Failed to put in plasma: {e}")
+                self.connect()
+                time.sleep(1)
+
+    def delete(self, name):
+        object_id = plasma.ObjectID(hashlib.sha1(str.encode(name)).digest())
+        while True:
+            try:
+                self.plasma_client.delete([object_id])
+                return
+            except:
+                self.connect()
+                time.sleep(1)

frigate/video.py

@@ -5,22 +5,19 @@ import cv2
 import queue
 import threading
 import ctypes
+import pyarrow.plasma as plasma
 import multiprocessing as mp
 import subprocess as sp
 import numpy as np
-import hashlib
-import pyarrow.plasma as plasma
-import SharedArray as sa
 import copy
 import itertools
 import json
 from collections import defaultdict
-from frigate.util import draw_box_with_label, area, calculate_region, clipped, intersection_over_union, intersection, EventsPerSecond
+from frigate.util import draw_box_with_label, area, calculate_region, clipped, intersection_over_union, intersection, EventsPerSecond, listen, PlasmaManager
 from frigate.objects import ObjectTracker
 from frigate.edgetpu import RemoteObjectDetector
 from frigate.motion import MotionDetector
 
-# TODO: add back opencv fallback
 def get_frame_shape(source):
     ffprobe_cmd = " ".join([
         'ffprobe',
@@ -99,16 +96,82 @@ def create_tensor_input(frame, region):
     # Expand dimensions since the model expects images to have shape: [1, 300, 300, 3]
     return np.expand_dims(cropped_frame, axis=0)
 
-def track_camera(name, config, ffmpeg_global_config, global_objects_config, detect_lock, detect_ready, frame_ready, detected_objects_queue, fps, skipped_fps, detection_fps):
+def start_or_restart_ffmpeg(ffmpeg_cmd, frame_size, ffmpeg_process=None):
-    print(f"Starting process for {name}: {os.getpid()}")
+    if not ffmpeg_process is None:
+        print("Terminating the existing ffmpeg process...")
+        ffmpeg_process.terminate()
+        try:
+            print("Waiting for ffmpeg to exit gracefully...")
+            ffmpeg_process.communicate(timeout=30)
+        except sp.TimeoutExpired:
+            print("FFmpeg didnt exit. Force killing...")
+            ffmpeg_process.kill()
+            ffmpeg_process.communicate()
+        ffmpeg_process = None
 
-    # Merge the ffmpeg config with the global config
+    print("Creating ffmpeg process...")
-    ffmpeg = config.get('ffmpeg', {})
+    print(" ".join(ffmpeg_cmd))
-    ffmpeg_input = get_ffmpeg_input(ffmpeg['input'])
+    process = sp.Popen(ffmpeg_cmd, stdout = sp.PIPE, stdin = sp.DEVNULL, bufsize=frame_size*10, start_new_session=True)
-    ffmpeg_global_args = ffmpeg.get('global_args', ffmpeg_global_config['global_args'])
+    return process
-    ffmpeg_hwaccel_args = ffmpeg.get('hwaccel_args', ffmpeg_global_config['hwaccel_args'])
+
-    ffmpeg_input_args = ffmpeg.get('input_args', ffmpeg_global_config['input_args'])
+class CameraCapture(threading.Thread):
-    ffmpeg_output_args = ffmpeg.get('output_args', ffmpeg_global_config['output_args'])
+    def __init__(self, name, ffmpeg_process, frame_shape, frame_queue, take_frame, fps, detection_frame):
+        threading.Thread.__init__(self)
+        self.name = name
+        self.frame_shape = frame_shape
+        self.frame_size = frame_shape[0] * frame_shape[1] * frame_shape[2]
+        self.frame_queue = frame_queue
+        self.take_frame = take_frame
+        self.fps = fps
+        self.skipped_fps = EventsPerSecond()
+        self.plasma_client = PlasmaManager()
+        self.ffmpeg_process = ffmpeg_process
+        self.current_frame = 0
+        self.last_frame = 0
+        self.detection_frame = detection_frame
+
+    def run(self):
+        frame_num = 0
+        self.skipped_fps.start()
+        while True:
+            if self.ffmpeg_process.poll() != None:
+                print(f"{self.name}: ffmpeg process is not running. exiting capture thread...")
+                break
+
+            frame_bytes = self.ffmpeg_process.stdout.read(self.frame_size)
+            self.current_frame = datetime.datetime.now().timestamp()
+
+            if len(frame_bytes) == 0:
+                print(f"{self.name}: ffmpeg didnt return a frame. something is wrong.")
+                continue
+
+            self.fps.update()
+
+            frame_num += 1
+            if (frame_num % self.take_frame) != 0:
+                self.skipped_fps.update()
+                continue
+
+            # if the detection process is more than 1 second behind, skip this frame
+            if self.detection_frame.value > 0.0 and (self.last_frame - self.detection_frame.value) > 1:
+                self.skipped_fps.update()
+                continue
+
+            # put the frame in the plasma store
+            self.plasma_client.put(f"{self.name}{self.current_frame}",
+                    np
+                        .frombuffer(frame_bytes, np.uint8)
+                        .reshape(self.frame_shape)
+                )
+            # add to the queue
+            self.frame_queue.put(self.current_frame)
+            self.last_frame = self.current_frame
+
+def track_camera(name, config, global_objects_config, frame_queue, frame_shape, detection_queue, detected_objects_queue, fps, detection_fps, read_start, detection_frame):
+    print(f"Starting process for {name}: {os.getpid()}")
+    listen()
+
+    detection_frame.value = 0.0
 
     # Merge the tracked object config with the global config
     camera_objects_config = config.get('objects', {})    
@@ -122,18 +185,7 @@ def track_camera(name, config, ffmpeg_global_config, global_objects_config, dete
     for obj in objects_with_config:
         object_filters[obj] = {**global_object_filters.get(obj, {}), **camera_object_filters.get(obj, {})}
 
-    expected_fps = config['fps']
+    frame = np.zeros(frame_shape, np.uint8)
-    take_frame = config.get('take_frame', 1)
-
-    frame_shape = get_frame_shape(ffmpeg_input)
-    frame_size = frame_shape[0] * frame_shape[1] * frame_shape[2]
-
-    try:
-        sa.delete(name)
-    except:
-        pass
-
-    frame = sa.create(name, shape=frame_shape, dtype=np.uint8)
 
     # load in the mask for object detection
     if 'mask' in config:
@@ -146,66 +198,36 @@ def track_camera(name, config, ffmpeg_global_config, global_objects_config, dete
         mask[:] = 255
 
     motion_detector = MotionDetector(frame_shape, mask, resize_factor=6)
-    object_detector = RemoteObjectDetector('/labelmap.txt', detect_lock, detect_ready, frame_ready)
+    object_detector = RemoteObjectDetector(name, '/labelmap.txt', detection_queue)
 
     object_tracker = ObjectTracker(10)
 
-    ffmpeg_cmd = (['ffmpeg'] +
+    plasma_client = PlasmaManager()
-            ffmpeg_global_args +
-            ffmpeg_hwaccel_args +
-            ffmpeg_input_args +
-            ['-i', ffmpeg_input] +
-            ffmpeg_output_args +
-            ['pipe:'])
-
-    print(" ".join(ffmpeg_cmd))
-    
-    ffmpeg_process = sp.Popen(ffmpeg_cmd, stdout = sp.PIPE, bufsize=frame_size)
-    
-    plasma_client = plasma.connect("/tmp/plasma")
-    frame_num = 0
     avg_wait = 0.0
     fps_tracker = EventsPerSecond()
-    skipped_fps_tracker = EventsPerSecond()
     fps_tracker.start()
-    skipped_fps_tracker.start()
     object_detector.fps.start()
     while True:
-        start = datetime.datetime.now().timestamp()
+        read_start.value = datetime.datetime.now().timestamp()
-        frame_bytes = ffmpeg_process.stdout.read(frame_size)
+        frame_time = frame_queue.get()
-        duration = datetime.datetime.now().timestamp()-start
+        duration = datetime.datetime.now().timestamp()-read_start.value
+        read_start.value = 0.0
         avg_wait = (avg_wait*99+duration)/100
+        detection_frame.value = frame_time
+        
+        # Get frame from plasma store
+        frame = plasma_client.get(f"{name}{frame_time}")
 
-        if not frame_bytes:
+        if frame is plasma.ObjectNotAvailable:
-            break
-
-        # limit frame rate
-        frame_num += 1
-        if (frame_num % take_frame) != 0:
             continue
 
         fps_tracker.update()
         fps.value = fps_tracker.eps()
         detection_fps.value = object_detector.fps.eps()
-
-        frame_time = datetime.datetime.now().timestamp()
-        
-        # Store frame in numpy array
-        frame[:] = (np
-                    .frombuffer(frame_bytes, np.uint8)
-                    .reshape(frame_shape))
         
         # look for motion
         motion_boxes = motion_detector.detect(frame)
 
-        # skip object detection if we are below the min_fps and wait time is less than half the average
-        if frame_num > 100 and fps.value < expected_fps-1 and duration < 0.5*avg_wait:
-            skipped_fps_tracker.update()
-            skipped_fps.value = skipped_fps_tracker.eps()
-            continue
-        
-        skipped_fps.value = skipped_fps_tracker.eps()
-
         tracked_objects = object_tracker.tracked_objects.values()
 
         # merge areas of motion that intersect with a known tracked object into a single area to look at
@@ -305,7 +327,7 @@ def track_camera(name, config, ffmpeg_global_config, global_objects_config, dete
 
                 for index in idxs:
                     obj = group[index[0]]
-                    if clipped(obj, frame_shape): #obj['clipped']:
+                    if clipped(obj, frame_shape):
                         box = obj[2]
                         # calculate a new region that will hopefully get the entire object
                         region = calculate_region(frame_shape, 
@@ -345,8 +367,7 @@ def track_camera(name, config, ffmpeg_global_config, global_objects_config, dete
         # now that we have refined our detections, we need to track objects
         object_tracker.match_and_update(frame_time, detections)
 
-        # put the frame in the plasma store
-        object_id = hashlib.sha1(str.encode(f"{name}{frame_time}")).digest()
-        plasma_client.put(frame, plasma.ObjectID(object_id))
         # add to the queue
         detected_objects_queue.put((name, frame_time, object_tracker.tracked_objects))
+
+    print(f"{name}: exiting subprocess")