1.先用labelme 标注数据集

安装参考链接
https://www.pianshen.com/article/8141190515/

2.将生成的json进行相应的转换（批处理）

https://blog.csdn.net/gongzi123/article/details/86669036
其中：[python][原创]labelme_json_to_dataset没有生成info.yaml解决方法
https://blog.csdn.net/FL1623863129/article/details/112317861

第二种方法
https://github.com/wkentaro/labelme/tree/master/examples/semantic_segmentation

3.no named module skimage
解决办法：
https://jingyan.baidu.com/article/fdffd1f8673187f3e98ca133.html

https://www.pythonheidong.com/blog/article/430584/a36ace49fb1feb5da174/

4, original_keras_version = f.attrs[‘keras_version’].decode(‘utf8’) AttributeError: ‘str’ object has no attribute 'decode
卸载原来的h5py模块，安装2.10版本
https://blog.csdn.net/qq_41251963/article/details/110132101
5.pycocotools
直接pip 安装就可以

进行测试：
https://blog.csdn.net/u012746060/article/details/82143285
https://blog.csdn.net/qq_29462849/article/details/81037343