Quick Start

Quickly run the YOLOv5 inference demo to verify the NPU and become familiar with the basic workflow.

Download the demo package

Run the following command to download the RK3576 YOLOv5 demo archive.

radxa@rock-4d$

wget https://dl.radxa.com/rock4/4d/images/rk3576_rknn_yolov5_demo.tar.gz

If the download is slow, download the file on a PC first and transfer it to the device via scp.

Extract the demo project

Extract the archive to obtain the rk3576_rknn_yolov5_demo directory, which contains the model, test image, and executable.

radxa@rock-4d$

tar -xzf rk3576_rknn_yolov5_demo.tar.gz

Run the inference demo

Enter the directory and run the executable. The sample command uses bus.jpg as the input image.

radxa@rock-4d$

cd rk3576_rknn_yolov5_demo
./rknn_yolov5_demo ./model/yolov5s_relu_rk3576.rknn ./model/bus.jpg

Command parameters:

• ./rknn_yolov5_demo: Prebuilt RKNN inference sample
• ./model/yolov5s_relu_rk3576.rknn: Quantized YOLOv5s model tailored for RK3576
• ./model/bus.jpg: Sample input image

Check the results

After the inference succeeds, the terminal prints model information, parsed results, and the confidence for each detection, for example:

load label ./model/coco_80_labels_list.txt
model input num: 1, output num: 3
input tensors:
  index=0, name=images, n_dims=4, dims=[1, 640, 640, 3], n_elems=1228800, size=1228800, fmt=NHWC, type=INT8, qnt_type=AFFINE, zp=-128, scale=0.003922
output tensors:
  index=0, name=output0, n_dims=4, dims=[1, 255, 80, 80], n_elems=1632000, size=1632000, fmt=NCHW, type=INT8, qnt_type=AFFINE, zp=-128, scale=0.003922
  index=1, name=286, n_dims=4, dims=[1, 255, 40, 40], n_elems=408000, size=408000, fmt=NCHW, type=INT8, qnt_type=AFFINE, zp=-128, scale=0.003922
  index=2, name=288, n_dims=4, dims=[1, 255, 20, 20], n_elems=102000, size=102000, fmt=NCHW, type=INT8, qnt_type=AFFINE, zp=-128, scale=0.003922
model is NHWC input fmt
model input height=640, width=640, channel=3
origin size=640x640 crop size=640x640
input image: 640 x 640, subsampling: 4:2:0, colorspace: YCbCr, orientation: 1
scale=1.000000 dst_box=(0 0 639 639) allow_slight_change=1 _left_offset=0 _top_offset=0 padding_w=0 padding_h=0
rga_api version 1.10.1_[0]
rknn_run
person @ (209 243 286 510) 0.880
person @ (479 238 560 526) 0.871
person @ (109 237 232 534) 0.832
bus @ (93 129 553 464) 0.705
person @ (79 353 122 517) 0.301
write_image path: out.png width=640 height=640 channel=3 data=0x892c940

• person @ (...) 0.880 shows the detected class, bounding box, and confidence score.
• write_image path: out.png indicates that the bounding boxes were drawn and saved to out.png.

The current directory retains the inference output out.png: