# linkta-yolov8-train

**Repository Path**: flamebox/linkta-yolov8-train

## Basic Information

- **Project Name**: linkta-yolov8-train
- **Description**: No description available
- **Primary Language**: Python
- **License**: MulanPSL-2.0
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2024-08-09
- **Last Updated**: 2024-08-09

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

## 1. 构建环境
确认当前为root账号，执行
```
chmod +x ./build_env.sh && ./build_env.sh
```
如果在安装torch时网速太慢，可以停止自动构建脚本，手动下载以下两个文件
```
https://download.pytorch.org/whl/cu118/torch-2.1.2%2Bcu118-cp310-cp310-linux_x86_64.whl
https://download.pytorch.org/whl/cu118/torchvision-0.16.2%2Bcu118-cp310-cp310-linux_x86_64.whl
```
再次进入虚拟环境,顺序执行以下命令，手工完成环境安装
```
source .yolov8_venv/bin/activate
pip install -r torch-2.1.2+cu118-cp310-cp310-linux_x86_64.whl
pip install -r torchvision-0.16.2+cu118-cp310-cp310-linux_x86_64.whl
pip install openvino==2024.0.0
pip install nncf
pip install ultralytics==8.0.170
pip install opencv-python-headless
cp files/Arial.ttf /root/.config/Ultralytics/
```
## 2. 训练和导出模型
- 终端输入以下命令，进入虚拟环境
```c
  source .yolov8_venv/bin/activate
```
- 修改train.yaml,指定训练数据集路径和分类名称。数据集路径需要使用**绝对路径**
```sh
# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..]
path: /train_data/pig # dataset root dir
train: train/images # train images (relative to 'path') 118287 images
val: valid/images # val images (relative to 'path') 5000 images
test: test-dev2017.txt # 20288 of 40670 images, submit to https://competitions.codalab.org/competitions/20794

# Classes
names:
  0: pig
```
- 开始训练
```
# 避免多卡训练报错
export MKL_SERVICE_FORCE_INTEL=1

yolo detect train data=train.yaml model=yolov8s.yaml epochs=300 batch=24 imgsz=640 device=0
```
- 导出模型
  将训练好的`best.pt`文件复制到`my_models`
```
cd my_models
python yolov8_openvino_int8.py best.pt ../train.yaml
```
- 量化脚本会转换模型、对比评分、对比测试推理速度，最后的结果模型会保存在`my_models/xxx_openvino_model`中。
`xxx_quantized.xml`和`xxx_quantized.bin`就是给天露部署的模型