Object Detection
Object Detection is a core task in computer vision, primarily used to identify and locate target objects (such as people, vehicles, and animals) in images or video, typically by annotating target positions with bounding boxes. It is widely used in security surveillance, autonomous driving, industrial inspection, and similar scenarios. Representative model families include YOLO, R-CNN, and DETR.
| Hardware Requirements | Model | Performance Benchmark |
|---|---|---|
| RDK S600 USB Camera | YOLO26x | 120 fps (single core) |
Hardware Connection
Connect the USB camera to the USB port of the RDK S600 development board.
Model Deployment
Environment Setup
# This case uses a single-core model
# Download yolo26x_demo.tar (also available via wget), place on the board, and extract
wget https://archive.d-robotics.cc/downloads/kol_test/yolo26x_demo.tar
tar xvf yolo26x_demo.tar
Launch the Case
# Enter the target path
cd yolo26x_demo/ultralytics_yolo26/runtime/python
# Specify an image for detection; specify the image via --test-img and the model via --model-path
python3 main.py --task detect --model-path ../../model/yolo26x_nashp_640x640_nv12.hbm --label-file ../../../coco/coco_classes.names --test-img kite.jpg --img-save-path result.jpg
# Start the camera; specify the USB camera via --camera-id
# To ensure proper and smooth visualization functionality, it is recommended to connect a screen via HDMI and run the command in a terminal on the desktop
python3 main.py --camera-id 0 --task detect --model-path ../../model/yolo26x_nashp_640x640_nv12.hbm --label-file ../../../coco/coco_classes.names