Zoom-out Features For Image Understanding
I will describe a novel feed-forward architecture, which maps small image elements (pixels or superpixels) to rich feature representations extracted from a sequence of nested regions of increasing extent. These regions are obtained by "zooming out" from the superpixel all the way to scene-level resolution. Applied to semantic segmentation, our approach exploits statistical structure in the image and in the label space without setting up explicit structured prediction mechanisms, and thus avoids complex and expensive inference. Instead superpixels are classified by a feedforward multilayer network with skip-layer connections spanning the zoomout levels. Using off-the-shelf network, pre-trained on ImageNet classification task, this zoom-out architecture achieves near state-of-the-art accuracy on the PASCAL VOC 2012 test set. Joint work with Mohammadreza Mostajabi and Payman Yadollahpour.
Since February 2008, I am an Assistant Professor at TTI-Chicago, a philanthropically endowed academic computer science institute located on the University of Chicago campus. I also hold a part-time faculty appointment at the University of Chicago Department of Computer Science. Prior to coming to TTI-Chicago, I was a post-doctoral researcher at the Department of Computer Science of Brown University where I worked with Michael Black. I received my PhD degree at MIT where I worked at CSAIL with Trevor Darrell on computer vision and machine learning. My thesis topic was Learning Task-Specific Similarity. Before coming to MIT, I was a graduate student in the Computer Science Department of the Technion, Israel Institute of Technology in Haifa, Israel, where I got my MSc thesis under the advisement of Ran El-Yaniv and Yoram Baram. I got my undergraduate degree in Math and CS from Hebrew University in Jerusalem, Israel.