YOLO11-POSE

This version of YOLO11-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_yolo11_pose
The repo of axcl-samples, which you can get the how to build the axcl_yolo11_pose
Pulsar2 Link, How to Convert ONNX to axmodel

Support Platform

AX650
- M4N-Dock(爱芯派Pro)
- M.2 Accelerator card
AX630C

Chips	cost
AX650	25 ms
AX630C	TBD ms

How to use

Download all files from this repository to the device

(axcl) axera@raspberrypi:~/samples/AXERA-TECH/YOLO11-Pose $ tree -L 2
.
├── ax620e
│   └── yolo11s-pose.axmodel
├── ax650
│   └── yolo11x-pose.axmodel
├── ax_aarch64
│   └── ax_yolo11_pose
├── axcl_aarch64
│   └── axcl_yolo11_pose
├── axcl_x86_64
│   └── axcl_yolo11_pose
├── config.json
├── football.jpg
├── README.md
├── yolo11_pose_config.json
├── yolo11_pose_out.jpg
├── yolo11s-pose-cut.onnx
└── yolo11s-pose.onnx

6 directories, 12 files

Inference

Input image:

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

root@ax650:~/YOLO11-Pose# ./ax_aarch64/ax_yolo11_pose -m ax650/yolo11x-pose.axmodel -i football.jpg
--------------------------------------
model file : ax650/yolo11x-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:1.40 ms
--------------------------------------
Repeat 1 times, avg time 25.21 ms, max_time 25.21 ms, min_time 25.21 ms
--------------------------------------
detection num: 6
 0:  94%, [1350,  337, 1632, 1036], person
 0:  93%, [ 492,  477,  658, 1000], person
 0:  92%, [ 756,  219, 1126, 1154], person
 0:  91%, [   0,  354,  314, 1108], person
 0:  73%, [   0,  530,   81, 1017], person
 0:  54%, [ 142,  589,  239, 1013], person
--------------------------------------

Output image:

Inference with M.2 Accelerator card

(axcl) axera@raspberrypi:~/samples/AXERA-TECH/YOLO11-Pose $ chmod +x axcl_aarch64/axcl_yolo11_pose
(axcl) axera@raspberrypi:~/samples/AXERA-TECH/YOLO11-Pose $ ./axcl_aarch64/axcl_yolo11_pose -m ax650/yolo11x-pose.axmodel -i football.jpg
--------------------------------------
model file : ax650/yolo11x-pose.axmodel
image file : football.jpg
img_h, img_w : 640 640
--------------------------------------
axclrtEngineCreateContextt is done.
axclrtEngineGetIOInfo is done.

grpid: 0

input size: 1
    name:   images
        1 x 640 x 640 x 3


output size: 6
    name: /model.23/Concat_1_output_0
        1 x 80 x 80 x 65

    name: /model.23/Concat_2_output_0
        1 x 40 x 40 x 65

    name: /model.23/Concat_3_output_0
        1 x 20 x 20 x 65

    name: /model.23/cv4.0/cv4.0.2/Conv_output_0
        1 x 80 x 80 x 51

    name: /model.23/cv4.1/cv4.1.2/Conv_output_0
        1 x 40 x 40 x 51

    name: /model.23/cv4.2/cv4.2.2/Conv_output_0
        1 x 20 x 20 x 51

==================================================

Engine push input is done.
--------------------------------------
post process cost time:0.43 ms
--------------------------------------
Repeat 1 times, avg time 25.05 ms, max_time 25.05 ms, min_time 25.05 ms
--------------------------------------
detection num: 6
 0:  94%, [1350,  337, 1632, 1036], person
 0:  93%, [ 492,  477,  658, 1000], person
 0:  92%, [ 756,  219, 1126, 1154], person
 0:  91%, [   0,  354,  314, 1108], person
 0:  73%, [   0,  530,   81, 1017], person
 0:  54%, [ 142,  589,  239, 1013], person
--------------------------------------

AXERA-TECH
/

YOLO11-Pose