图像处理之阴影检测实例BDRAR

paper :Bidirectional Feature Pyramid Network with Recurrent Attention Residual Modules for Shadow Detection

github:BDRAR

Preparation

1. Set the path of pretrained ResNeXt model in resnext/config.py
2. Set the path of SBU dataset in config.py

The pretrained ResNeXt model is ported from the official torch version, using the convertor provided by clcarwin. You can directly download the pretrained model ported by me.

 

 

效果展示：

以下左边是true shadow mask，右图是predict mask，下面是GT

good:

 

not good :

1.本身是类似阴影的黑色物体被误判为阴影，当然这可能我训练的模型效果有关

 

2.漏检部分：

 

整体来讲，效果还是阔以的。