Spatial Transformers and Deformable CNNs
演讲者： Dai Jifeng (Microsoft Research Asia)
Abstract: A key challenge in visual recognition is how to accommodate geometric variations or model geometric transformations in object scale, pose, viewpoint, and part deformation. Meanwhile, CNNs are inherently limited to model large, unknown transformations. The limitation originates from the fixed geometric structures of CNN modules. In this talk, we would introduce Spatial Transformer Networks and Deformable ConvNets. They seek to learn spatial transformation from data in a deep learning framework, and are representative in the field.
Bio: Jifeng Dai is currently a Lead Researcher in Visual Computing Group at Microsoft Research Asia (MSRA). Before he joined MSRA in 2014, he received the B.S. degree and the Ph.D. degree from Tsinghua University with honor in 2009 and 2014 respectively. He was also a visiting student to University of California, Los Angeles (UCLA) from 2012 to 2013. His current research focus is on deep learning for high-level vision, especially for semantic segmentation and object detection. He is the first author of R-FCN for object detection and Deformable ConvNets. He and his team members has won COCO challenge in 2015 and 2016.