congjianting

通常情况下是训练数据存在些问题, 比如越界啥的.

不过目前用该方法还没有得到一个满意的效果, 暂时换别的方法了.

检查下训练数据, 包括类别,和坐标位置.

请问有可用的训练代码参考吗?

@kbrajwani can you share the edited codes for trainning only on word level? i think many people will benefit from it.

@ML-Mr-J 请问内存泄漏问题解决了吗?

@cwjhx 非常感谢

训练数据格式是否就是个标准的coco数据格式?

by the way, the model structure is cascade-res50-fpn-maskrcnn.

@seasonyang i face the same problem, the training loss is very large, did you know why and tell me how to fix it?