NanoDet-RepVGG-a1-640 on reComputer R Series (CM5 + Hailo-8)

nanodet_repvgg_a1_640 performs COCO 80-class object detection on Hailo-8 through HailoRT. Same base (base/nanodet.yaml) as nanodet_repvgg but with a larger RepVGG-A1 backbone and 640x640 input for higher accuracy. NMS runs on-chip (Hailo HPP, meta_arch=yolov8), so the host only parses the post-NMS tensor.

This page targets reComputer R Series (CM5 + Hailo-8) with a PCIe Hailo-8 accelerator.

Model Info

The accuracy value is Hailo's Model Zoo reference. It is not a benchmark measured on CM5.

Hardware and Host Setup

Install the Hailo-8 packages on the host and confirm that the driver and device are ready before starting the container:

sudo apt update
sudo apt install hailort hailort-pcie-driver python3-hailort
sudo reboot

# After reboot
hailortcli --version
hailortcli fw-control identify
ls -l /dev/hailo0

The expected HailoRT version for this image is 4.23.x.

Run With Demo Video

sudo docker run --rm \
  --name cm5-hailo8-nanodet-repvgg-a1-640 \
  --privileged \
  --net=host \
  -e PYTHONUNBUFFERED=1 \
  --device /dev/hailo0:/dev/hailo0 \
  -v /usr/lib/libhailort.so.4.23.0:/usr/lib/libhailort.so.4.23.0:ro \
  -v /usr/lib/libhailort.so:/usr/lib/libhailort.so:ro \
  ghcr.io/seeed-projects/recomputer-hailo8-cv/nanodet_repvgg_a1_640:latest \
  python web_detection.py --model_path model/nanodet_repvgg_a1_640.hef --video_path video/test.mp4

Open http://<Board_IP>:8000 to view the web preview.

USB Camera Mode

sudo docker run --rm \
  --name cm5-hailo8-nanodet-repvgg-a1-640 \
  --privileged \
  --net=host \
  -e PYTHONUNBUFFERED=1 \
  --device /dev/hailo0:/dev/hailo0 \
  --device /dev/video0:/dev/video0 \
  -v /usr/lib/libhailort.so.4.23.0:/usr/lib/libhailort.so.4.23.0:ro \
  -v /usr/lib/libhailort.so:/usr/lib/libhailort.so:ro \
  ghcr.io/seeed-projects/recomputer-hailo8-cv/nanodet_repvgg_a1_640:latest \
  python web_detection.py --model_path model/nanodet_repvgg_a1_640.hef --camera_id 0

REST API

Prediction endpoint:

POST http://<Board_IP>:8000/api/models/nanodet_repvgg_a1_640/predict

Example image request:

curl -X POST "http://<Board_IP>:8000/api/models/nanodet_repvgg_a1_640/predict" \
  -F "file=@test.jpg"

Implementation Notes

• Same base (base/nanodet.yaml) and on-chip NMS (HPP) post-processing as nanodet_repvgg; only the backbone (RepVGG-A1) and input size (640x640) differ.
• The HEF runs NMS on-chip (device_pre_post_layers: nms=true, hpp=true, meta_arch=yolov8); the app only parses the post-NMS tensor, so nms_thresh is ignored (API parity only).
• Post-NMS rows are [ymin, xmin, ymax, xmax, score], normalized to [0,1] of the 640x640 letterboxed input; the app scales to pixels and un-letterboxes.
• normalize_in_net + on-chip input_conversion(bgr_to_rgb): the app letterboxes with black (0) padding and feeds raw uint8 BGR pixels — no manual normalization, no cvtColor.
• HailoRT returns the NMS vstream as a ragged per-class list (NMS-by-score); the parser handles that plus object/dense layouts. First inference logs the raw type/shape so it can be verified on hardware.
• Class mapping: cls_id (0..79) → standard COCO 80-class list directly.
• Defaults: confidence 0.25, IOU 0.45 (IOU has no effect — NMS on-chip). Model Zoo eval uses 0.05 / 0.6 — lower the confidence slider to inspect more.

Development Notes

• Source module: src/rpi5_hailo8_nanodet_repvgg_a1_640/
• Dockerfile: docker/hailo8/nanodet_repvgg_a1_640.dockerfile
• Container: ghcr.io/seeed-projects/recomputer-hailo8-cv/nanodet_repvgg_a1_640:latest
• The module uses the HailoRT 4.23 VStreams API and a Hailo-8-specific HEF.
• Family: nanodet (variants nanodet_repvgg / nanodet_repvgg_a12 / nanodet_repvgg_a1_640); this is the nanodet_repvgg_a1_640 build.