AXERA-TECH/YOLOv8-Pose

YOLOv8-POSE

This version of YOLOv8-POSE has been converted to run on the Axera NPU using w8a16 quantization.

This model has been optimized with the following LoRA:

Compatible with Pulsar2 version: 3.4

Convert tools links:

For those who are interested in model conversion, you can try to export axmodel through

• The repo of ax-samples, which you can get the how to build the ax_yolov8_pose

• The repo of axcl-samples, which you can get the how to build the axcl_yolov8_pose

• Pulsar2 Link, How to Convert ONNX to axmodel

Support Platform

• AX650
  • M4N-Dock(爱芯派Pro)
  • M.2 Accelerator card
• AX630C
  • 爱芯派2
  • Module-LLM
  • LLM630 Compute Kit

How to use

Download all files from this repository to the device

root@ax650 ~/yolov8-pose # tree -L 2
.
├── ax650
│   └── yolov8s-pose.axmodel
├── ax_aarch64
│   └── ax_yolov8_pose
├── config.json
├── football.jpg
├── README.md
├── yolov8_pose_config.json
├── yolov8_pose_out.jpg
├── yolov8s-pose-cut.onnx
└── yolov8s-pose.onnx

3 directories, 9 files

Inference

Input image:

Inference with AX650 Host, such as M4N-Dock(爱芯派Pro)

root@ax650 ~/yolov8-pose # ./ax_yolov8_pose -m yolov8s-pose.axmodel -i football.jpg
--------------------------------------
model file : yolov8s-pose.axmodel
image file : football.jpg
img_h, img_w : 640 640
--------------------------------------
Engine creating handle is done.
Engine creating context is done.
Engine get io info is done.
Engine alloc io is done.
Engine push input is done.
--------------------------------------
post process cost time:1.24 ms
--------------------------------------
Repeat 1 times, avg time 10.97 ms, max_time 10.97 ms, min_time 10.97 ms
--------------------------------------
detection num: 4
 0:  93%, [ 760,  211, 1125, 1157], person
 0:  93%, [1349,  337, 1633, 1039], person
 0:  92%, [   0,  354,  324, 1104], person
 0:  88%, [ 489,  474,  656,  996], person
--------------------------------------

Output image: