# GearYOLO

**Repository Path**: tommy0x7c00/gear-yolo

## Basic Information

- **Project Name**: GearYOLO
- **Description**: 基于 Geartrain 的YOLO
- **Primary Language**: Python
- **License**: Apache-2.0
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2024-05-23
- **Last Updated**: 2024-09-09

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

<p align="center">
    <br>
    <img src="assets/Geartrain.jpg" width="1200"/>
    <br>
</p>

# gearyolo

## 介绍

基于 `Geartrain` 的目标检测模型集合

## 快速开始

```bash
 python3 examples/yolov8_test.py --img_path assets/zidane.jpg
```

<p align="left">
    <br>
    <img src="assets/result.jpg" width="200"/>
    <br>
</p>

## 任务
| [**YOLOv8**][通用目标检测](gearyolo/yolov8/README.md) | [**Damo-YOLO**][火焰烟雾检测](gearyolo/damo_yolo/README.md) |
| :---------------------------------------------------: | :---------------------------------------------------------: |
|   <img src='assets/detection.jpg' height="126px" >  <img src='assets/keypoint.jpg' height="126px">  <img src='assets/segment.jpg' height="126px">   |   <img src='assets/smokefire_result.jpg' height="126px">    |


## Benchmark

<table style="width:100%; border: 1px solid black; border-collapse: collapse;">
  <tr>
    <th style="border: 1px solid black; padding: 8px;text-align:center; vertical-align:middle;">Model</th>
    <th style="border: 1px solid black; padding: 8px;text-align:center; vertical-align:middle;">Framework</th>
    <th style="border: 1px solid black; padding: 8px;text-align:center; vertical-align:middle;">Device</th>
    <th style="border: 1px solid black; padding: 8px;text-align:center; vertical-align:middle;">Time (ms)</th>
  </tr>
  <tr>
    <td rowspan="2"><a href="models/yolov8n.onnx">YOLOv8n</td>
    <td rowspan="2" style="border: 1px solid black; padding: 8px;">onnxruntime (fp32)</td>
    <td style="border: 1px solid black; padding: 8px;">CPU (Intel(R) Xeon(R) Gold 5118 CPU @ 2.30GHz)</td>
    <td style="border: 1px solid black; padding: 8px;">180.22</td>
    

  </tr>
  <tr>
    <td style="border: 1px solid black; padding: 8px;">CUDA (NVIDIA GeForce RTX 2080)</td>
    <td style="border: 1px solid black; padding: 8px;">9.31</td>
  </tr>

  <tr>
    <td rowspan="2"><a href="models/yolov8n.onnx">DAMO YOLO tinynasL25_S</td>
    <td rowspan="2" style="border: 1px solid black; padding: 8px;">onnxruntime (fp32)</td>
    <td style="border: 1px solid black; padding: 8px;">CPU (Intel(R) Xeon(R) Gold 5118 CPU @ 2.30GHz)</td>
    <td style="border: 1px solid black; padding: 8px;">215.07</td>
    

  </tr>
  <tr>
    <td style="border: 1px solid black; padding: 8px;">CUDA (NVIDIA GeForce RTX 2080)</td>
    <td style="border: 1px solid black; padding: 8px;">17.92</td>
  </tr>
</table>



## TritonInferenceServer
### download
```bash
cd /workspace
git clone https://gitee.com/tommy0x7c00/gear-cv.git
git clone https://gitee.com/tommy0x7c00/gear-train-py.git
git clone https://gitee.com/tommy0x7c00/gear-yolo.git
```
```bash
export PYTHONPATH=$PYTHONPATH:\
/workspace/gear-cv:\
/workspace/gear-train-py:\
/workspace/gear-yolo
```

### export onnx model
```
yolo export model=yolov8n.pt format=onnx opset=16
```


### quick_dev
- run image
```bash
docker run -it --rm --shm-size 100g --name test \
-v /data/workspace/tangm/tommy0x7c00/gear-yolo/triton_models:/models \
-v /data/workspace/tangm/tommy0x7c00/gear-cv:/workspace/gear-cv \
-v /data/workspace/tangm/tommy0x7c00/gear-yolo:/workspace/gear-yolo \
-v /data/workspace/tangm/tommy0x7c00/gear-train-py:/workspace/gear-train-py \
-p 8900:8000 \
-p 8901:8001 \
-p 8902:8002 \
bywin.harbor.com:52/nvidia/tritonserver-gear2-dev:22.08-py3 /bin/bash
```
- run server
```bash
export PYTHONPATH=/workspace/gear-cv:/workspace/gear-train-py:/workspace/gear-yolo:${PYTHONPATH}

tritonserver --model-repository /models
```

### quick_deploy
- build image
```bash
docker build -t bywin.harbor.com:52/aisp/tritonserver-gear2-yolo:22.08-py3-v1.0.0 -f docker/ort-gpu-runtime.dockerfile ..
```
- run image

```bash
docker run -it --rm --shm-size 100g --name test \
-v /data/workspace/tangm/tommy0x7c00/gear-yolo/triton_models/object_detection:/models/object_detection \
-p 8900:8000 \
-p 8901:8001 \
-p 8902:8002 \
bywin.harbor.com:52/aisp/tritonserver-gear2-yolo:22.08-py3-v1.0.0
```


### test client
```bash
pip install requests grpcio tritonclient -i https://pypi.tuna.tsinghua.edu.cn/simple
```

```bash
python object_detect_client.py
```

## [RKNN 模型格式部署](gearyolo/rk_yolov8/README.md)