README.md

---
license: mit
language:
- en
pipeline_tag: object-detection
tags:
- YOLOv7
- YOLOv7-Face
---

# YOLOv7-FACE

This version of YOLOv7-FACE 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 AXera Platform](https://github.com/AXERA-TECH/ax-samples), which you can get the detial of guide

- [Pulsar2 Link, How to Convert ONNX to axmodel](https://pulsar2-docs.readthedocs.io/en/latest/pulsar2/introduction.html) 


## Support Platform

- AX650
  - [M4N-Dock(爱芯派Pro)](https://wiki.sipeed.com/hardware/zh/maixIV/m4ndock/m4ndock.html)
  - [M.2 Accelerator card](https://axcl-docs.readthedocs.io/zh-cn/latest/doc_guide_hardware.html)
- AX630C
  - [爱芯派2](https://axera-pi-2-docs-cn.readthedocs.io/zh-cn/latest/index.html)
  - [Module-LLM](https://docs.m5stack.com/zh_CN/module/Module-LLM)
  - [LLM630 Compute Kit](https://docs.m5stack.com/zh_CN/core/LLM630%20Compute%20Kit)
  
|Chips|cost|
|--|--|
|AX650| 12.6 ms |
|AX630C| TBD ms |

## How to use

Download all files from this repository to the device

```
root@ax650:~/YOLOv7-Face# tree
.
|-- ax650
|   `-- yolov7-face.axmodel
|-- ax_yolov7_face
|-- selfie.jpg
`-- yolov7_face_out.jpg
```

### Inference

Input image:
![](./selfie.jpg)

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

```
root@ax650:~/YOLOv7-Face# ./ax_yolov7_face -m ax650/yolov7-face.axmodel -i selfie.jpg
--------------------------------------
model file : ax650/yolov7-face.axmodel
image file : selfie.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:8.70 ms
--------------------------------------
Repeat 1 times, avg time 12.59 ms, max_time 12.59 ms, min_time 12.59 ms
--------------------------------------
detection num: 174
 0:  91%, [1137,  869, 1283, 1065], face
 0:  91%, [1424,  753, 1570,  949], face
......
 0:  45%, [1658,  362, 1677,  387], face
 0:  45%, [1445,  437, 1467,  462], face
--------------------------------------
root@ax650:~/YOLOv7-Face#
```

Output image:
![](./yolov7_face_out.jpg)


#### Inference with M.2 Accelerator card
```
(base) axera@raspberrypi:~/lhj/YOLOv7-Face $ ./axcl_aarch64/axcl_yolov7_face -m ax650/yolov7-face.axmodel -i selfie.jpg 
--------------------------------------
model file : ax650/yolov7-face.axmodel
image file : selfie.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: 3
    name:      511 
        1 x 80 x 80 x 63

    name:      520 
        1 x 40 x 40 x 63

    name:      529 
        1 x 20 x 20 x 63

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

Engine push input is done. 
--------------------------------------
post process cost time:8.29 ms 
--------------------------------------
Repeat 1 times, avg time 12.23 ms, max_time 12.23 ms, min_time 12.23 ms
--------------------------------------
detection num: 277
 0:  91%, [1137,  869, 1283, 1065], face
 0:  91%, [1424,  753, 1570,  949], face
 0:  89%, [1305,  764, 1403,  900], face
 0:  87%, [1738,  786, 1796,  860], face
 ......
 0:  20%, [1120,  570, 1145,  604], face
 0:  20%, [1025,  390, 1041,  413], face
--------------------------------------
```
Output image:
![](./yolov7_face_axcl_out.jpg)