# pytorch-yolo

**Repository Path**: ecustdmf/pytorch-yolo

## Basic Information

- **Project Name**: pytorch-yolo
- **Description**: 使用PyTorch框架进行YOLO系列模型的训练，包括YOLOv3系列、YOLOv5系列等，用于DMF液滴检测。
- **Primary Language**: Unknown
- **License**: GPL-3.0
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 3
- **Forks**: 1
- **Created**: 2021-06-02
- **Last Updated**: 2021-09-07

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# YOLO PyTorch版本
使用PyTorch框架进行YOLO系列模型的训练，包括YOLOv3系列、YOLOv5系列等，用于DMF液滴检测。训练框架源于 YOLOv5: <https://github.com/ultralytics/yolov5> 


# 教程

## 安装依赖
```
pip install -r requirements.txt
```

## 配置项目结构
按以下文件结构配置数据集和Python工程
```
├─HLDROPLIB    
│  ├─train   
│  │  ├─images  
│  │  └─labels  
│  └─val  
│      ├─images  
│      └─labels  
└─pytorch-yolo  
    ├─data  
    ├─models  
    ├─utils 
    ├─...
    └─weights  
```

## 训练
按实际硬件环境配置参数，如：

- 使用GPU训练 YOLOv5s:`python  train.py  --name yolov5s  --cfg yolov5s.yaml  --epochs 100  --batch-size 4  --workers 4  --device 0`
- 从断点恢复训练：`python  train.py  --resume`
- 使用Tensorboard对训练过程进行可视化：<http://localhost:6006/> `tensorboard --logdir runs` 
- 使用CPU训练 YOLOv3-tiny:`python  train.py  --name yolov3-tiny  --cfg yolov3-tiny.yaml  --epochs 100  --batch-size 8  --workers 4  --device cpu`
![train_log](https://gitee.com/ecustdmf/pytorch-yolo/raw/master/data/snapshots/train_log.png)

## 测试
默认测试路径 `./data/images`，输出路径 `./runs/detect/$name`
```
python  detect.py  --name yolov3-tiny  --weights runs/train/yolov3-tiny/weights/best.pt  --conf-thres 0.6  --iou-thres 0.2
```
![detect_log](https://gitee.com/ecustdmf/pytorch-yolo/raw/master/data/snapshots/detect_log.png)
![detect_results](https://gitee.com/ecustdmf/pytorch-yolo/raw/master/data/snapshots/detect_results.jpg)
## 导出
将模型导出为 `torchscript` 以在C++环境下进行部署，导出为 `ONNX` 以进行网络结构可视化(<https://github.com/lutzroeder/netron>)
```
python  ./models/export.py  --weights runs/train/yolov3-tiny/weights/best.pt  --img 640  --batch 1  --train  --device 0
```
![export_log](https://gitee.com/ecustdmf/pytorch-yolo/raw/master/data/snapshots/export_log.png)

## 其他
参见YOLOv5相关教程：<https://github.com/ultralytics/yolov5>
> * [Train Custom Data](https://github.com/ultralytics/yolov5/wiki/Train-Custom-Data)
> * [Tips for Best Training Results](https://github.com/ultralytics/yolov5/wiki/Tips-for-Best-Training-Results)
> * [Weights & Biases Logging](https://github.com/ultralytics/yolov5/issues/1289)
> * [Supervisely Ecosystem](https://github.com/ultralytics/yolov5/issues/2518)
> * [Multi-GPU Training](https://github.com/ultralytics/yolov5/issues/475)
> * [PyTorch Hub](https://github.com/ultralytics/yolov5/issues/36)
> * [TorchScript, ONNX, CoreML Export](https://github.com/ultralytics/yolov5/issues/251) 
> * [Test-Time Augmentation (TTA)](https://github.com/ultralytics/yolov5/issues/303)
> * [Model Ensembling](https://github.com/ultralytics/yolov5/issues/318)
> * [Model Pruning/Sparsity](https://github.com/ultralytics/yolov5/issues/304)
> * [Hyperparameter Evolution](https://github.com/ultralytics/yolov5/issues/607)
> * [Transfer Learning with Frozen Layers](https://github.com/ultralytics/yolov5/issues/1314)
> * [TensorRT Deployment](https://github.com/wang-xinyu/tensorrtx)

# 致谢
- ultralytics/yolov5: <https://github.com/ultralytics/yolov5>
- WongKinYiu/PyTorch_YOLOv4: <https://github.com/WongKinYiu/PyTorch_YOLOv4>