CVPR Daily - Wednesday

Emanuel Laude is a PhD student in the Computer Vision Group at the Technical University of Munich under the supervision of Daniel Cremers. He spoke to us at his poster yesterday. Emanuel tells us that they provide a general-purpose-solver for semi- supervised and transductive learning problems. In contrast to semi- supervised learning, in transductive learning there is no training phase. Instead the labels are inferred directly at test-time incorporating the given labelled training examples. He says the task can be formulated as a hard non-convex mixed-integer energy minimization problem . With this formulation, for instance, they tackle the problem of interactive image segmentation and supervised video object segmentation. The task can be formulated as a hard non-convex mixed-integer energy minimization problem. 8 Emanuel Laude Wednesday From left: Laura Leal-Taixé, Emanuel Laude and Jan-Hendrik Lange in front of their poster