To address the issue of false positive and false negative error predictions in shadow detection tasks,a shadow detection network based on error prediction processing,abbreviated as EPSD,is proposed.The EPSD network improves the recognition rate of false negative pixels through the error prediction process-ing module,and strengthens the inspection of false positive pixels,which achieves enhanced recognition of both types of error prediction phenomena,thereby improving the overall detection performance of the shadow detection network.Experiment results show that compared to existing shadow detection networks,the EPSD network achieves superior performance on the SBU dataset.
关键词
深度学习/卷积神经网络/阴影检测
Key words
deep learning/Convolutional Neural Network/shadow detection
维普
万方数据