中央研究院 資訊科學研究所

This talks will cover two works: SSD: Single Shot Multibox Detector and DSSD : DeConvolutional Single Shot Detector. SSD discretizes the output space of bounding boxes into a set of default boxes over different aspect ratios and scales per feature map location. At prediction time, the network generates scores for the presence of each object category in each default box and produces adjustments to the box to better match the object shape. Additionally, the network combines predictions from multiple feature maps with different resolutions to naturally handle objects of various sizes. SSD is simple relative to methods that require object proposals because it completely eliminates proposal generation and subsequent pixel or feature resampling stages and encapsulates all computation in a single network. In order to introduce additional context into state-of-the-art general object detection, we first combine a state-of-the-art classifier Residual-101 with a fast detection framework SSD, and then augment SSD+Residual-101 with deconvolutional layers to introduce additional large-scale context in object detection and improve accuracy, especially for small objects, calling our resulting system DSSD for deconvolutional single shot detector. While these two contributions are easily described at a high-level, a naïve implementation does not succeed. Instead we show that carefully adding additional stages of learned transformations, specifically a module for feed-forward connections in deconvolution and a new output module, enables this new approach and forms a potential way forward for further detection research.

Cheng-Yang Fu is a third-year Ph.D. student at the University of North Carolina at Chapel Hill in the Department of Computer Science and advised by Prof. Alex C. Berg in the Computer Vision Group now. In 2010, he received B.S. and M.S. in Computer Science from National Tsing Hua University under the supervision of Prof. Ren-Song Tsay working on multi-core simulation and embedded systems. Before studying his Ph.D. program, he worked as a research assistant at Academia Sinica in Taiwan with Dr. Tyng-Luh Liu and started the interests on computer vision and deep learning. His research works include all kinds of problems of computer vision and deep learning, but especially focus on object detection.