Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/doubleZ0108/Data-Augmentation
General Data Augmentation Algorithms for Object Detection(esp. Yolo)
https://github.com/doubleZ0108/Data-Augmentation
data-augmentation dataset object-detection yolo
Last synced: about 1 month ago
JSON representation
General Data Augmentation Algorithms for Object Detection(esp. Yolo)
- Host: GitHub
- URL: https://github.com/doubleZ0108/Data-Augmentation
- Owner: doubleZ0108
- Created: 2020-07-28T15:29:49.000Z (over 4 years ago)
- Default Branch: master
- Last Pushed: 2023-04-14T03:50:12.000Z (over 1 year ago)
- Last Synced: 2024-08-03T01:20:55.875Z (4 months ago)
- Topics: data-augmentation, dataset, object-detection, yolo
- Language: Python
- Homepage:
- Size: 537 KB
- Stars: 12
- Watchers: 2
- Forks: 4
- Open Issues: 3
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
- awesome-object-detection-datasets - doubleZ0108/Data-Augmentation - Augmentation?style=social"/> : General Data Augmentation Algorithms for Object Detection(esp. Yolo). (Summary)
- awesome-yolo-object-detection - doubleZ0108/Data-Augmentation - Augmentation?style=social"/> : General Data Augmentation Algorithms for Object Detection(esp. Yolo). (Extensional Frameworks)
- awesome-yolo-object-detection - doubleZ0108/Data-Augmentation - Augmentation?style=social"/> : General Data Augmentation Algorithms for Object Detection(esp. Yolo). (Extensional Frameworks)
README
# 数据集扩充
[中文(zh-cn)](https://github.com/doubleZ0108/Data-Augmentation/blob/master/README.md) | [English(en)](https://github.com/doubleZ0108/Data-Augmentation/blob/master/README_en.md)
通用目标检测深度学习数据集扩充方法(尤其针对Yolo)
* [具体实现](#具体实现)
* [如何运行](#如何运行)
* [开发环境](#开发环境)
* [关于作者](#关于作者)
------
## 具体实现
[数据集扩充方法详细说明](https://github.com/doubleZ0108/Data-Augmentation/blob/master/doc/data_augmentation_detailed.md)
- 图像强度变换
- 亮度变化： [lightness](https://github.com/doubleZ0108/Data-Augmentation/blob/master/src/lightness.py) [darkness](https://github.com/doubleZ0108/Data-Augmentation/blob/master/src/darkness.py)
- 对比度变化：[contrast](https://github.com/doubleZ0108/Data-Augmentation/blob/master/src/contrast.py)
- 图像滤波
- 锐化：[sharpen](https://github.com/doubleZ0108/Data-Augmentation/blob/master/src/sharpen.py)
- 高斯模糊：[blur](https://github.com/doubleZ0108/Data-Augmentation/blob/master/src/blur.py)
- 透视变换
- 镜像翻转：[flip](https://github.com/doubleZ0108/Data-Augmentation/blob/master/src/flip.py)
- 图像裁剪：[crop](https://github.com/doubleZ0108/Data-Augmentation/blob/master/src/crop.py)
- 图像拉伸：[deform](https://github.com/doubleZ0108/Data-Augmentation/blob/master/src/deform.py)
- 镜头畸变：[distortion](https://github.com/doubleZ0108/Data-Augmentation/blob/master/src/distortion.py)
- 注入噪声
- 椒盐噪声：[noise](https://github.com/doubleZ0108/Data-Augmentation/blob/master/src/noise.py)
- 渐晕：[vignetting](https://github.com/doubleZ0108/Data-Augmentation/blob/master/src/vignetting.py)
- 其他
- 随机抠除：[cutout](https://github.com/doubleZ0108/Data-Augmentation/blob/master/src/cutout.py)
> 按从左到右，从上到下顺序排列
## 如何运行
- 安装所需依赖
```bash
pip install -r requirements.txt
```
- 将原始图片和[标注数据](https://github.com/doubleZ0108/IDEA-Lab-Summer-Camp/blob/master/doc/Study-Notes/labelImg工具.md)放到`data/`中
- 运行
- `python main.py` 生成所有数据在`data/`文件夹下
- 执行对应`.py`文件生成指定扩充图像数据
- 使用
```python
os.chdir(dirname)
(name, appidx) = os.path.splitext(filename)
img = np.array(Image.open(filename)) # Image读入的图片形式
somealgo_img = somealgo(np.copy(img)) # 复制一份传入扩充算法
somealgo_img.save(name + "_somealgo" + appidx) # 将扩充图像写入本地
saveSomeAlgoLabel(name) # 自动生成对应的标注(部分算法需要手动标注)
```
## 开发环境
- **操作系统**: macOS Catalina 10.15.5
- **开发语言**: python 3.7.4
- **主要依赖**: numpy | cv2 | PIL | shutil
## 关于作者
| 姓名👤 | 张喆 \| doubleZ |
| :-----------: | :-------------------------------------------------: |
| **学校🏫** | 同济大学 |
| **联系方式✉️** | [[email protected]](mailto:[email protected]) |