[Detector Support]: orange pi 5B never detects anything

[carlhung]

Describe the problem you are having

unable to detect anything on my orange pi 5B

My orange pi 5B is running on:
Distributor ID: Debian
Description: Armbian 24.2.1 bookworm
Release: 12
Codename: bookworm

The kernel is:
5.10.160-legacy-rk35xx

Version

0.13.2-6476F8A

Frigate config file

mqtt:
  enabled: False

detectors: # required
  rknn: # required
    type: rknn # required
    # core mask for npu
    core_mask: 0b111

model: # required
  # name of yolov8 model or path to your own .rknn model file
  # possible values are:
  # - default-yolov8n
  # - default-yolov8s
  # - default-yolov8m
  # - default-yolov8l
  # - default-yolov8x
  # - /config/model_cache/rknn/your_custom_model.rknn
  path: default-yolov8x
  # width and height of detection frames
  width: 320
  height: 320
  # pixel format of detection frame
  # default value is rgb but yolov models usually use bgr format
  input_pixel_format: bgr # required
  # shape of detection frame
  input_tensor: nhwc

cameras:
  Front_Door: # <--- this will be changed to your actual camera later
    enabled: True
    ffmpeg:
      inputs:
        - path: rtsp:https://frontdoor:[email protected]:554/live/ch0
          roles:
            - detect
            - record
      hwaccel_args: preset-rk-h265
    detect:
        enabled: True
        width: 320
        height: 320

docker-compose file or Docker CLI command

version: "3.9"
services:
  frigate:
    container_name: frigate
    privileged: true
    restart: unless-stopped
    image: ghcr.io/blakeblackshear/frigate:stable-rk
    shm_size: "512mb" 
    volumes:
      - /etc/localtime:/etc/localtime:ro
      - $HOME/externalDisk/frigate/config:/config
      - $HOME/externalDisk/frigate/storage:/media/frigate
      - type: tmpfs 
        target: /tmp/cache
        tmpfs:
          size: 1000000000
    ports:
      - "5000:5000"
      - "8554:8554"
      - "8555:8555/tcp"
      - "8555:8555/udp"
    environment:
      FRIGATE_RTSP_PASSWORD: "xxxxx"

Relevant log output

2024-03-21 14:48:12.796544721  [INFO] Preparing Frigate...
2024-03-21 14:48:12.842436105  [INFO] Starting Frigate...
2024-03-21 14:48:14.545143967  [2024-03-21 14:48:14] frigate.app                    INFO    : Starting Frigate (0.13.2-6476f8a)
2024-03-21 14:48:14.562930144  [2024-03-21 14:48:14] peewee_migrate.logs            INFO    : Starting migrations
2024-03-21 14:48:14.567965410  [2024-03-21 14:48:14] peewee_migrate.logs            INFO    : There is nothing to migrate
2024-03-21 14:48:14.576083259  [2024-03-21 14:48:14] frigate.app                    INFO    : Recording process started: 454
2024-03-21 14:48:14.578920846  [2024-03-21 14:48:14] frigate.app                    INFO    : go2rtc process pid: 93
2024-03-21 14:48:14.603277310  [2024-03-21 14:48:14] detector.rknn                  INFO    : Starting detection process: 463
2024-03-21 14:48:14.609721599  [2024-03-21 14:48:14] frigate.app                    INFO    : Output process started: 465
2024-03-21 14:48:14.623883283  [2024-03-21 14:48:14] frigate.app                    INFO    : Camera processor started for Front_Door: 470
2024-03-21 14:48:14.632598166  [2024-03-21 14:48:14] frigate.app                    INFO    : Capture process started for Front_Door: 473
2024-03-21 14:48:14.931431149  I RKNN: [14:48:14.931] RKNN Runtime Information: librknnrt version: 1.5.2 (c6b7b351a@2023-08-23T15:28:22)
2024-03-21 14:48:14.931457107  I RKNN: [14:48:14.931] RKNN Driver Information: version: 0.9.2
2024-03-21 14:48:14.931923476  I RKNN: [14:48:14.931] RKNN Model Information: version: 6, toolkit version: 1.5.2+b642f30c(compiler version: 1.5.2 (c6b7b351a@2023-08-23T07:39:01)), target: RKNPU v2, target platform: rk3588, framework name: ONNX, framework layout: NCHW, model inference type: static_shape

Operating system

Debian

Install method

Docker Compose

Coral version

Other

Any other information that may be helpful

No response

[NickM-27]

CC @MarcA711

[MarcA711]

Hey,
could you try to remove the model cache folder (it is inside the config folder)? If you have some custom models saved inside, create a backup.

cd /path/to/frigate/config
sudo rm -rf model_cache/

[carlhung]

I removed it and restarted the container. it still doesn't work. I saw it recreated a file of the model.

2024-03-22 21:12:52.463596443  [INFO] Preparing Frigate...
2024-03-22 21:12:52.506313943  [INFO] Starting Frigate...
2024-03-22 21:12:54.248027233  [2024-03-22 21:12:54] frigate.app                    INFO    : Starting Frigate (0.13.2-6476f8a)
2024-03-22 21:12:54.248308104  [2024-03-22 21:12:54] frigate.app                    INFO    : Creating directory: /config/model_cache
2024-03-22 21:12:54.268963947  [2024-03-22 21:12:54] peewee_migrate.logs            INFO    : Starting migrations
2024-03-22 21:12:54.274143585  [2024-03-22 21:12:54] peewee_migrate.logs            INFO    : There is nothing to migrate
2024-03-22 21:12:54.311672697  [2024-03-22 21:12:54] frigate.app                    INFO    : Recording process started: 453
2024-03-22 21:12:54.322983376  [2024-03-22 21:12:54] frigate.app                    INFO    : go2rtc process pid: 89
2024-03-22 21:12:54.393175209  [2024-03-22 21:12:54] detector.rknn                  INFO    : Starting detection process: 462
2024-03-22 21:12:54.399538413  [2024-03-22 21:12:54] frigate.detectors.plugins.rknn INFO    : Downloading yolov8x model.
2024-03-22 21:12:54.407375390  [2024-03-22 21:12:54] frigate.app                    INFO    : Output process started: 464
2024-03-22 21:12:54.460262292  [2024-03-22 21:12:54] frigate.app                    INFO    : Camera processor started for Front_Door: 469
2024-03-22 21:12:54.461326861  [2024-03-22 21:12:54] frigate.app                    INFO    : Capture process started for Front_Door: 472
2024-03-22 21:13:32.036913589  I RKNN: [21:13:32.036] RKNN Runtime Information: librknnrt version: 1.5.2 (c6b7b351a@2023-08-23T15:28:22)
2024-03-22 21:13:32.036939547  I RKNN: [21:13:32.036] RKNN Driver Information: version: 0.9.2
2024-03-22 21:13:32.037319292  I RKNN: [21:13:32.037] RKNN Model Information: version: 6, toolkit version: 1.5.2+b642f30c(compiler version: 1.5.2 (c6b7b351a@2023-08-23T07:39:01)), target: RKNPU v2, target platform: rk3588, framework name: ONNX, framework layout: NCHW, model inference type: static_shape

[MarcA711]

Could you try my current development version:

cd ~
git clone -b 13.2 https://github.com/MarcA711/frigate-rockchip
cd frigate-rockchip
sudo make local-rk

Afterwards, change the image in your docker-compose.yml to image: frigate:latest-rk

[carlhung]

sorry, I dont really know how to use it.

[carlhung]

sorry, just found out. after removing the model from the config folder and restarted. objection detection works now. thanks

[MarcA711]

Great! If you need further assistance, please contact me.