应002资讯网宋明辉、张达治、郭志昌老师邀请，香港浸会大学数学系台雪成教授作学术报告，欢迎感兴趣的师生参加！

【报告题目】：Deep Neural Networks for star-shape and convex shape representations

【报告时间】：2020年7月6日，下午14：00

【报告平台】：Zoom

点击链接直接加入会议：

https://tudelft.zoom.us/j/99740856333?pwd=QlhwMVNyTFlYSHNxdE9tYUh4MEh6dz09

【会议】：997 4085 6333

【会议密码】：3vRmNt

【报告摘要】：Convex Shapes (CS) are common priors for optic disc and cup segmentation in eye fundus images. It is important to design proper techniques to represent convex shapes. So far, it is still a problem to guarantee that the output objects from a Deep Neural Convolution Networks (DCNN) are convex shapes. In this work, we propose a technique which can be easily integrated into the commonly used DCNNs for image segmentation and guarantee that outputs are convex shapes. This method is flexible and it can handle multiple objects and allow some of the objects to be convex. Our method is based on the dual representation of the sigmoid activation function in DCNNs. In the dual space, the convex shape prior can be guaranteed by a simple quadratic constraint on a binary representation of the shapes. Moreover, our method can also integrate spatial regularization and some other shape prior using a soft thresholding dynamics (STD) method. The regularization can make the boundary curves of the segmentation objects to be simultaneously smooth and convex. We design a very stable active set projection al- gorithm to numerically solve our model. This algorithm can form a new plug-and-play DCNN layer called CS-STD whose outputs must be a nearly binary segmentation of convex objects. In the CS-STD block, the convexity information can be propagated to guide the DCNN in both forward and backward propagation during training and prediction process. As an application example, we apply the convexity prior layer to the retinal fundus images segmentation by taking the popular DeepLabV3+ as a backbone network. Experimental results on several public datasets show that our method is efficient and outperforms the classical DCNN segmentation methods.

【报告人简介】：Prof Tai Xue-Cheng is currently a full professor at the Department of Mathematics, Hong Kong Baptist University. He received his Bachelor degree in Mathematical Science from Zhengzhou University, Licenciate and Ph.D. degrees from the University of Jyvaskyla. His research interests include Numerical PDEs, optimization techniques, inverse problems, and image processing. He has done significant research work in his research areas and published many research works in top quality international conference and journals. He served as organizing and program committee members for a number of international conferences and has been often invited speakers for international conferences. He has served as referee and reviewers for many premier conferences and journals. Dr. Tai is a member of the editor board for several internatiuonal journals.