Frigate with Raspberry Pi 5 AI Kit and Hailo 8L - Troubleshooting Help Needed

Hello community,

I’m trying to set up Frigate with the Raspberry Pi 5 and a Hailo 8L AI accelerator. I’m running into a few issues, and I’m hoping someone can offer guidance or share any solutions if they’ve encountered similar problems.

Current Setup

  • Hardware: Raspberry Pi 5, Hailo 8L AI accelerator
  • Software: Frigate 0.15.0, Docker
  • Goal: Use the Hailo 8L for object detection with Frigate.

Problem Summary

Frigate starts successfully, but when initializing the Hailo device for detection, it encounters the following errors and then stops.

Error Log Highlights

Below are the key error messages:

  1. VDevice Not Defined: When the Hailo plugin initializes, I see the error: NameError: name ‘VDevice’ is not defined
    This suggests an issue with recognizing VDevice within the Hailo SDK.

  2. HailoRTException Not Defined: Shortly afterward, Frigate encounters: NameError: name ‘HailoRTException’ is not defined

  3. which may indicate an incomplete or improper setup of the Hailo SDK.

  4. Detection Stalls: The Frigate watchdog notes “Detection appears to have stopped,” resulting in Frigate restarting.

  5. Resource Warnings: Upon shutdown, there’s a message about leaked shared memory objects: resource_tracker: There appear to be 50 leaked shared_memory objects to clean up at shutdown

Steps Taken

  • Installed the Hailo SDK in a virtual environment due to the managed environment restriction.
  • Bound the virtual environment directory to Docker (if necessary) to make the SDK accessible to Frigate.
  • Verified device recognition via lspci, confirming the Hailo device is detected.

Questions

  • Has anyone successfully run Frigate with the Hailo 8L on a Raspberry Pi setup?
  • Is there any specific Docker configuration required to support Hailo SDK dependencies within the container?
  • Any insights on handling the VDevice and HailoRTException issues?

Any help or suggestions would be greatly appreciated. Thanks in advance!

1 Like

Hey @bostan_89,

Welcome to the Hailo Community!

I’ve run Frigate with RPi5 and Hailo8L before. Can you confirm which Docker image you’re using? Make sure it ends with ‘h8l’ - this indicates it includes HailoRT.

Could you also share the specific error message you’re encountering? This would help me better assist you with troubleshooting.

Best regards,
Omri

Now I’m using image: ghcr.io/blakeblackshear/frigate:0.15.0-beta2-h8l

Input tag ‘hailo8l’ found using ‘type’ does not match any of the expected tags: ‘cpu’, ‘deepstack’, ‘edgetpu’, ‘onnx’, ‘openvino’, ‘rknn’, ‘rocm’, 'tens
2024-11-21 16:29:48.194503158 *************************************************************
2024-11-21 16:29:48.194507213 *** End Config Validation Errors

1 Like

So I had beta 1 running well but had the same issue I tried to upgrade to beta 2, apparently it’s because beta requires RT 4.19 but that isn’t out yet for the pi 5, that’s what I found on the beta 2 thread in GitHub.

Hi. Did you get it to work. I have similar issues and ready to stick everything in the trash and bet Google Home.
Please let me know

Hey all,

Can you please each specify the problem he is facing ?
As for the docker image , you can choose the latest with h8l for hailo. Not only the beta.

Please provide more details so i can help you

Having the exact same issue on the latest beta version with latest 14.9 hailo drivers. Beta version 1 works but requires 14.7 driver.

Hello,

I having same issue, Frigate is installed as Home Assistant Addon Frigate Beta
below is the config.yaml
name: Frigate Beta
version: 0.15.0-beta4
breaking_versions:

  • 0.15.0-beta4
  • 0.15.0-beta3
  • 0.15.0-beta2
  • 0.15.0-beta1
  • 0.14.1
  • 0.14.0
    panel_icon: “mdi:cctv”
    panel_title: Frigate
    slug: frigate-beta
    description: NVR with realtime local object detection for IP cameras
    url: “GitHub - blakeblackshear/frigate: NVR with realtime local object detection for IP cameras
    image: ghcr.io/blakeblackshear/frigate:0.15.0-beta4-h8l
    shm_size: “512mb” # update for your cameras based on calculation above
    startup: application
    boot: auto
    init: false
    webui: “http://[HOST]:[PORT:5000]/”
    watchdog: “http://[HOST]:[PORT:5000]/”
    ingress: true
    ingress_port: 5000
    ingress_entry: /
    panel_admin: false
    homeassistant_api: true
    hassio_api: true
    ports:
    8555/tcp: 8555
    8555/udp: 8555
    8554/tcp: null
    8971/tcp: null
    5000/tcp: null
    1984/tcp: null
    1935/tcp: null
    ports_description:
    8555/tcp: WebRTC over tcp
    8555/udp: WebRTC over udp
    8554/tcp: RTSP Restream
    8971/tcp: Authenticated Web interface
    5000/tcp: Unauthenticated Web interface (Not required for Hass.io Ingress)
    1984/tcp: go2rtc API
    1935/tcp: RTMP streams
    host_network: false
    devices:
  • /dev/dri/renderD128
  • /dev/apex_0
  • /dev/apex_1
  • /dev/apex_2
  • /dev/apex_3
  • /dev/dri/card0
  • /dev/vchiq
  • /dev/video10
  • /dev/video0
  • /dev/hailo0
    usb: true
    video: true
    tmpfs: true
    full_access: false
    privileged:
  • PERFMON
    environment:
    CONFIG_FILE: /config/frigate.yml
    schema:
    plus_api_key: password?
    services:
  • “mqtt:want”
    arch:
  • amd64
  • aarch64
    map:
  • “media:rw”
  • “config:rw”
    Frigate.yaml config
    detectors:
    hailo8l:
    type: hailo8l
    device: PCIe
    model:
    path: /config/model_cache/h8l_cache/ssd_mobilenet_v1.hef
    model:
    width: 300
    height: 300
    input_tensor: nhwc
    input_pixel_format: bgr
    model_type: ssd

log when start the Frigate Addon
2025-01-07 06:54:40.761315732 [2025-01-07 13:54:40] detector.hailo8l INFO : Starting detection process: 398
2025-01-07 06:54:40.767653885 Process detector:hailo8l:
2025-01-07 06:54:40.767663737 Traceback (most recent call last):
2025-01-07 06:54:40.767665126 File “/opt/frigate/frigate/detectors/plugins/hailo8l.py”, line 72, in init
2025-01-07 06:54:40.767672904 self.target = VDevice()
2025-01-07 06:54:40.767674533 NameError: name ‘VDevice’ is not defined
2025-01-07 06:54:40.767675200
2025-01-07 06:54:40.767676274 During handling of the above exception, another exception occurred:
2025-01-07 06:54:40.767679478
2025-01-07 06:54:40.767681867 Traceback (most recent call last):
2025-01-07 06:54:40.767683607 File “/usr/lib/python3.9/multiprocessing/process.py”, line 315, in _bootstrap
2025-01-07 06:54:40.767684348 self.run()
2025-01-07 06:54:40.767685478 File “/opt/frigate/frigate/util/process.py”, line 41, in run_wrapper
2025-01-07 06:54:40.767686311 return run(*args, **kwargs)
2025-01-07 06:54:40.767687385 File “/usr/lib/python3.9/multiprocessing/process.py”, line 108, in run
2025-01-07 06:54:40.767690255 self._target(*self._args, **self._kwargs)
2025-01-07 06:54:40.767691330 File “/opt/frigate/frigate/object_detection.py”, line 121, in run_detector
2025-01-07 06:54:40.767693237 object_detector = LocalObjectDetector(detector_config=detector_config)
2025-01-07 06:54:40.767694293 File “/opt/frigate/frigate/object_detection.py”, line 68, in init
2025-01-07 06:54:40.767695256 self.detect_api = create_detector(detector_config)
2025-01-07 06:54:40.767696589 File “/opt/frigate/frigate/detectors/init.py”, line 18, in create_detector
2025-01-07 06:54:40.767697385 return api(detector_config)
2025-01-07 06:54:40.767698515 File “/opt/frigate/frigate/detectors/plugins/hailo8l.py”, line 105, in init
2025-01-07 06:54:40.767699293 except HailoRTException as e:
2025-01-07 06:54:40.767728534 NameError: name ‘HailoRTException’ is not defined
2025-01-07 06:54:40.768108758 [2025-01-07 13:54:40] frigate.detectors.plugins.hailo8l INFO : Initializing Hailo device as PCIe
2025-01-07 06:54:40.769253877 [2025-01-07 13:54:40] frigate.detectors.plugins.hailo8l INFO : A model file already exists at /config/model_cache/h8l_cache/ssd_mobilenet_v1.hef not downloading one.
2025-01-07 06:54:40.809510996 [2025-01-07 13:54:40] frigate.app INFO : Output process started: 414

notes: I am using AI kit + (not the M.2 type) can this the issue

Finally manage to make it work

the problem is on config.yaml for downloading the correct images should be like below

will upload performance later

1 Like

Had the same issue and moved to Frigate beta 15.1. This sorted the “Type” issue