[Detector Support]: Issues with Coral detection getting stuck

[bigmike613]

Describe the problem you are having

I am using the USB Coral TPU and since last week I have been getting "Detection appears to be stuck. Restarting detection process..." error messages sporadically. I have tried the same coral on 2 different PCs and have also bought a new USB cable and new Coral device to rule out hardware issues. I have the same results. the detection is always force killed and then restarts fine and runs for some time. I am using a Frigate+ model, but I have also tried to go back to a standard model for troubleshooting and get the same result.

Version

0.15.0-cea210d

Frigate config file

# yaml-language-server: $schema=http://frigate:5000/api/config/schema.json
model:
#TPU
  path: plus://<redacted>

mqtt:
  enabled: true
  host: 10.10.10.250
  user: mqtt101
  password: <redacted>
ffmpeg:
  hwaccel_args: preset-nvidia-h264

detectors:
  coral:
    type: edgetpu
    device: usb


go2rtc:
  streams:
    Backyard:
      - rtsp://<redacted>@10.10.20.24:554/axis-media/media.amp
    Front:
      - rtsp://<redacted>@10.10.20.22:554/axis-media/media.amp
    Side:
      - rtsp://<redacted>@10.10.20.21:554/axis-media/media.amp
    Garage:
      - rtsp://<redacted>@10.10.20.23:554/axis-media/media.amp

record:
  #sync_recordings: true
  enabled: true
  retain:
    days: 90

objects:
  track:
    - person
    - car
    - dog
    - cat
    - deer
    - license_plate
    - face
    - package
    - amazon
    - ups
    - fedex
    - amazon
  filters:
    dog:
      min_score: .7
      threshold: .9
    cat:
      min_score: .65
      threshold: .8
    face:
      min_score: .7
    package:
      min_score: .65
      threshold: .9
    license_plate:
      min_score: .6
    amazon:
      min_score: .75
    ups:
      min_score: .75
    fedex:
      min_score: .75
    person:
      min_score: .7
      threshold: .85
    car:
      min_area: 500
      max_area: 100000
      min_ratio: 0.5
      min_score: 0.7
      threshold: 0.85
timestamp_style:
  format: '%m/%d/%Y %H:%M:%S'
  color:
    red: 255
    green: 255
    blue: 255
  thickness: 2
  effect: shadow
#---------------------------------------------------------------------------------------------------------------------------------------
birdseye:
  enabled: true
  mode: continuous
cameras:
  Backyard: # <------ Name the camera
    ffmpeg:
      inputs:
        - path: rtsp://127.0.0.1:8554/Backyard
          input_args: preset-rtsp-restream
          roles:
            - detect
            - record
    detect:
      enabled: true  # <---- disable detection until you have a working camera feed
      width: 1920  # <---- update for your camera's resolution
      height: 1080  # <---- update for your camera's resolution
    snapshots:
      enabled: true
      timestamp: true
      bounding_box: true
      crop: true
      height: 500
      required_zones:
        - entire_yard
    record:
      enabled: true
      retain:
        days: 30
        mode: all
      alerts:
        retain:
          days: 90
        pre_capture: 15
        post_capture: 15
      detections:
        retain:
          days: 90
    motion:
      mask: 
        0.296,0.191,0.421,0.213,0.489,0.239,0.676,0.305,0.863,0.351,0.956,0.29,0.94,0.155,0.906,0,0,0,0,0.277,0.176,0.208
    zones:
      entire_yard:
        coordinates: 
          1,1,1,0,0.908,0,0.957,0.285,0.866,0.345,0.676,0.308,0.561,0.266,0.441,0.229,0.296,0.19,0,0.278,0,1
        inertia: 3
        loitering_time: 0
    objects:
      filters:
        car:
          mask:
            - 585,226,1297,329,1297,167,0,101,0,343
    review:
      alerts:
        required_zones:
          - entire_yard
  Front: # <------ Name the camera
    ffmpeg:
      inputs:
        - path: rtsp://127.0.0.1:8554/Front
          input_args: preset-rtsp-restream
          roles:
            - detect
            - record
    detect:
      enabled: true # <---- disable detection until you have a working camera feed
      width: 1920 # <---- update for your camera's resolution
      height: 1080 # <---- update for your camera's resolution
    zones:
      goodFront:
        coordinates: 
          0,1,0.752,1,0.865,0.709,0.927,0.444,0.953,0.255,0.841,0.185,0.828,0.155,0.701,0.133,0.62,0.133,0.357,0.191,0,0.436
        inertia: 3
        loitering_time: 0
    snapshots:
      enabled: true
      timestamp: true
      bounding_box: true
      crop: true
      height: 500
      required_zones:
        - goodFront
    motion:
      mask:
        - 0.833,0.177,0.956,0.261,0.986,0,0.968,0,0.83,0
        - 0.354,0.202,0.563,0.146,0.683,0.132,0.83,0.155,0.831,0,0,0,0,0.462,0.177,0.314
    objects:
      filters:
        car:
          mask:
            - 0.831,0.197,0.954,0.283,0.986,0,0.968,0,0.83,0
            - 0.353,0.23,0.563,0.146,0.688,0.137,0.886,0.16,0.871,0,0,0,0,0.462
        person:
          mask: 
            0.353,0.202,0.563,0.146,0.64,0.135,0.691,0.132,0.869,0.163,0.975,0.219,0.896,0,0,0,0,0.462,0.047,0.405,0.186,0.301,0.319,0.219
    record:
      enabled: true
      retain:
        days: 30
        mode: all
      alerts:
        retain:
          days: 90
        pre_capture: 15
        post_capture: 15
      detections:
        retain:
          days: 90
    review:
      alerts:
        required_zones: goodFront
      detections:
        required_zones: goodFront
  Side: # <------ Name the camera
    ffmpeg:
      inputs:
        - path: rtsp://127.0.0.1:8554/Side
          input_args: preset-rtsp-restream
          roles:
            - detect
            - record
    detect:
      enabled: true # <---- disable detection until you have a working camera feed
      width: 1920 # <---- update for your camera's resolution
      height: 1080 # <---- update for your camera's resolution
    snapshots:
      enabled: true
      timestamp: true
      bounding_box: true
      crop: true
      height: 500
      required_zones:
        - goodside
    motion:
      mask: 0.515,0.301,0.657,0.31,0.956,0.351,1,0,0,0,0,0.452,0.207,0.331
    record:
      enabled: true
      retain:
        days: 30
        mode: all
      alerts:
        retain:
          days: 90
        pre_capture: 15
        post_capture: 15
      detections:
        retain:
          days: 90
    zones:
      goodside:
        coordinates: 
          0.916,1,0.95,0.423,0.954,0.355,0.897,0.348,0.845,0.335,0.658,0.309,0.508,0.296,0.259,0.328,0.202,0.332,0,0.453,0,1
        inertia: 3
        loitering_time: 0
    objects:
      filters:
        car:
          mask: 0.513,0.301,0.657,0.308,0.956,0.351,1,0,0,0,0,0.483
    review:
      alerts:
        required_zones:
          - goodside
  Garage: # <------ Name the camera
    ffmpeg:
      inputs:
        - path: rtsp://127.0.0.1:8554/Garage
          input_args: preset-rtsp-restream
          roles:
            - detect
            - record
    detect:
      enabled: true # <---- disable detection until you have a working camera feed
      width: 1920 # <---- update for your camera's resolution
      height: 1080 # <---- update for your camera's resolution
    zones:
      garageGrasshouse:
        coordinates: 1920,1080,1920,597,1586,468,512,441,0,592,0,1080
      garageGrass:
        coordinates: 0.553,0.136,0.858,0.262,0.774,0,0.36,0,0.358,0.162
        inertia: 3
        loitering_time: 0
      driveway:
        coordinates: 1920,682,1920,452,1861,415,1733,322,1639,273,1484,244,1691,591
    snapshots:
      enabled: true
      timestamp: true
      bounding_box: true
      crop: true
      height: 500
      #required_zones:
      #  - driveway
      #  - garageGrass
      #  - garageGrasshouse
    record:
      enabled: true
      retain:
        days: 30
        mode: all
      alerts:
        retain:
          days: 90
        pre_capture: 15
        post_capture: 15
      detections:
        retain:
          days: 90
    objects:
      filters:
        car:
          mask: 0.884,0.552,1,0.402,1,0.081,0.342,0.06,0.227,0.485
    review:
      alerts:
        required_zones:
          - driveway
          - garageGrass
          - garageGrasshouse
    motion:
      threshold: 25
      contour_area: 10
      improve_contrast: true
      mask:
        - 0.183,0.084,0.281,0.037,0.31,0.078,0.309,0.164,0.306,0.238,0.295,0.273,0.267,0.307,0.243,0.339,0.211,0.343,0.186,0.31,0.166,0.296,0.155,0.24,0.158,0.157
        - 0.862,0,0.853,0.232,1,0.393,1,0
version: 0.15-1
camera_groups:
  Birdseye:
    order: 1
    icon: LuBird
    cameras: birdseye

docker-compose file or Docker CLI command

services:
  frigate:
    container_name: frigate
    privileged: true # this may not be necessary for all setups
    restart: unless-stopped
    image: ghcr.io/blakeblackshear/frigate:stable-tensorrt
    deploy:    
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1 
              capabilities: [gpu]
    shm_size: "308mb"
    devices:
      - /dev/bus/usb:/dev/bus/usb
      - /dev/dri/renderD128 
    volumes:
      - /etc/localtime:/etc/localtime:ro
      - /FRIGATE/config/config.yml:/config/config.yml
      - /mnt/video6:/media/frigate
      - /FRIGATE/config:/config
      - /FRIGATE/trt-models:/trt-models
      - type: tmpfs # Optional: 1GB of memory, reduces SSD/SD Card wear
        target: /tmp/cache
        tmpfs:
          size: 1000000000
    ports:
      - "5000:5000"
      - "8554:8554" # RTSP feeds
      - "8555:8555/tcp" # WebRTC over tcp
      - "8555:8555/udp" # WebRTC over udp
    environment:
      FRIGATE_RTSP_PASSWORD: "<redacted>"
      PLUS_API_KEY: "<redacted>"
      YOLO_MODELS: "yolov7-320"

Relevant Frigate log output

2025-03-06 05:12:17.571321046  [2025-03-06 05:12:14] frigate.detectors.plugins.edgetpu_tfl INFO    : Attempting to load TPU as usb
2025-03-06 05:12:17.576207783  [2025-03-06 05:12:17] frigate.detectors.plugins.edgetpu_tfl INFO    : TPU found
2025-03-06 05:12:20.392673954  [INFO] Starting go2rtc healthcheck service...
2025-03-06 06:35:44.989850461  [2025-03-06 06:35:44] frigate.watchdog               INFO    : Detection appears to be stuck. Restarting detection process...
2025-03-06 06:35:44.989855559  [2025-03-06 06:35:44] root                           INFO    : Waiting for detection process to exit gracefully...
2025-03-06 06:36:15.017444433  [2025-03-06 06:36:15] root                           INFO    : Detection process didn't exit. Force killing...
2025-03-06 06:36:15.034271283  [2025-03-06 06:36:15] root                           INFO    : Detection process has exited...
2025-03-06 06:36:15.085942775  [2025-03-06 06:36:15] detector.coral                 INFO    : Starting detection process: 28183
2025-03-06 06:36:18.062770153  [2025-03-06 06:36:15] frigate.detectors.plugins.edgetpu_tfl INFO    : Attempting to load TPU as usb
2025-03-06 06:36:18.068603388  [2025-03-06 06:36:18] frigate.detectors.plugins.edgetpu_tfl INFO    : TPU found
2025-03-06 06:54:15.071370304  [2025-03-06 06:54:15] frigate.watchdog               INFO    : Detection appears to be stuck. Restarting detection process...
2025-03-06 06:54:15.072040241  [2025-03-06 06:54:15] root                           INFO    : Waiting for detection process to exit gracefully...
2025-03-06 06:54:45.094820342  [2025-03-06 06:54:45] root                           INFO    : Detection process didn't exit. Force killing...
2025-03-06 06:54:45.119301954  [2025-03-06 06:54:45] root                           INFO    : Detection process has exited...
2025-03-06 06:54:45.147532271  [2025-03-06 06:54:45] detector.coral                 INFO    : Starting detection process: 34258
2025-03-06 06:54:47.909448361  [2025-03-06 06:54:45] frigate.detectors.plugins.edgetpu_tfl INFO    : Attempting to load TPU as usb
2025-03-06 06:54:47.914825274  [2025-03-06 06:54:47] frigate.detectors.plugins.edgetpu_tfl INFO    : TPU found

Install method

Docker Compose

Object Detector

Coral

Screenshots of the Frigate UI's System metrics pages

Any other information that may be helpful

No response

[hawkeye217]

This is usually a power related issue. Since you've already swapped the cable, the next step is to try a powered USB hub. Do you have other USB devices running on the same system?

[bigmike613]

dell optiplex 7050
32gb ram
Intel(R) Core(TM) i7-7700 CPU
Intel HD Graphics 630
Nvidia GeForce GT 710
1tb nvme SSD

Linux version 6.11.0-19-generic, Ubuntu 24.04.2 LTS

[bigmike613]

Bus 002 Device 005: ID 18d1:9302 is the coral (see google-coral/edgetpu#536)

2025-03-09 23:32:48.476166047  [Sun Mar  9 **09:13:27** 2025] usb 2-6.3: Device not responding to setup address.
2025-03-09 23:32:48.476167718  [Sun Mar  9 09:13:27 2025] usb 2-6.3: Device not responding to setup address.
2025-03-09 23:32:48.476169282  [Sun Mar  9 09:13:27 2025] usb 2-6.3: device not accepting address 5, error -71
2025-03-09 23:32:48.476180300  [Sun Mar  9 09:13:27 2025] usb 2-6.3: reset SuperSpeed USB device number 5 using xhci_hcd

looks like it stalled

this time is at the start of the container. there are no logs from dmesg at the time of the detector hang.

[hawkeye217]

Since you are using an i7 7th gen, you might consider using the OpenVINO detector instead of the Coral.

[bigmike613]

I tried that and it had my CPU usage very high, and fans were much louder. If it's my only option, I guess I'll do that but would prefer to use the Coral I invested in. Especially when it was working previously.

[hawkeye217]

You might not have configured the detector correctly, then. It shouldn't cause high CPU usage.

If Frigate didn't change, something else did. You've exhausted the options we usually recommend for the Coral.