Oran Shayer (EE, Technion)
Recent advances in deep learning and convolutional neural networks (CNNs) have had a profound impact on almost every computer vision task. However, generic (non-semantic) image segmentation is a notable exception despite it being one of the most fundamental and widely studied tasks in this field. In this talk, we revisit the generic segmentation task and propose Deep Generic Segmentation (DGS) -- a new deep learning approach combined with conditional random fields (CRFs). Our method differs significantly from previous popular segmentation methods and consists of three stages: a new pixel-wise representation learning scheme used for generic segmentation, a segment seed generation stage, and a CRF for the final processing stage. We tested our representations and segmentation method on BSDS500 and Pascal Context. We show that we are able to learn meaningful representations for the context of segmentations and that the representations themselves achieve state-of-the-art segment similarity scores. We did not achieve optimal results on the generic segmentation task, but present promising and competitive results using this method.
*MSc seminar under supervision of Prof. Micha Lindenbaum