Greg Shakhnarovich of the University of Chicago will present “Zoom-out Features For Image Understanding” in the Machine Learning and Friends lunch Thursday Oct. 29 at 1 pm (arrive at 12:45 for pizza). An abstract follows.

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.