数据科学| Deep learning for image restoration/reconstruction
Speaker: Dr.Ji Hui (Associate Professor, National University of Singapore)
Title: Deep learning for image restoration/reconstruction
Abstract:Image recovery/reconstruction is about recovering images of interest from either their degraded version or other measurements, which usually requires solving some challenging ill-posed inverse problem. In recent years, there has been an enduring effort on exploiting the potential of deep learning in solving many challenging image recovery problems. Different from classification, image recovery/reconstruction requires to learn a mapping from very high-dimensional space to itself. In this talk, we will discuss how to design deep neural network with strong connections to traditional optimization model and Bayesian inference. Such NN-based solutions not only provide state-of-the-art performance, but also has a good understanding and interpretation for it output. Several applications in medical imaging and digital photography will be demonstrated in this talk.