讲座人：Yang Gao (Ph.D. student at UC Berkeley)
讲座地点：复旦大学张江校区计算机楼405 (405 Computer Building, Zhangjiang Campus, Fudan University)
Robust perception-action models should be learned from training data with diverse visual appearances and realistic behaviors, yet current approaches to deep visuomotor policy learning have been generally limited to in-situ models learned from a single vehicle or a simulation environment. We advocate learning a generic vehicle motion model from large scale crowd-sourced video data, and develop an end-to-end trainable architecture for learning to predict a distribution over future vehicle egomotion from instantaneous monocular camera observations and previous vehicle state. Our model incorporates a novel FCN-LSTM architecture, which can be learned from large-scale crowd-sourced vehicle action data, and leverages available scene segmentation side tasks to improve performance under a privileged learning paradigm.
Brief Bio of Yang Gao:
Yang Gao is a Ph.D. student at University of California Berkeley under professor Trevor Darrell, which he joined in Fall 2014. Before that he was an undergraduate researcher in professor Jun Zhu’s group at Tsinghua University. He has also spent time working at Google research labs both in Beijing and Mountain View.
His research interests are computer vision, reinforcement learning and 3D scene reconstruction methods, with application emphasis on autonomous vehicles. He has also worked on multi-modal sensory fusion and fine-grained visual recognition problems.