IPAB Seminar 13/10/2017

Speaker: Olga Veksler

Title: Efficient Graph Cut Optimization for Full CRFs with Quantized Edges

Abstract: Fully connected pairwise Conditional Random Fields (Full-CRF) with Gaussian edge weights can achieve superior results compared to sparsely connected CRFs. However, traditional methods for Full-CRFs are too expensive. Previous work develops efficient approximate optimization based on mean field inference, which is a local optimization method and can be far from the optimum. We propose efficient and effective optimization based on graph cuts for Full-CRFs with {\em quantized} edge weights. To quantize edge weights, we partition the image into superpixels and assume that the weight of an edge between any two pixels depends only on the superpixels these pixels belong to. Our quantized edge CRF is an approximation to the Gaussian edge CRF, and gets closer to it as superpixel size decreases. Being an approximation, our model offers an intuition about the regularization properties of the Guassian edge Full-CRF. For efficient inference, we first consider the two-label case and develop an approximate method based on transforming the original problem into a smaller domain. Then we handle multi-label CRF by showing how to implement expansion moves. In both binary and multi-label cases, our solutions have significantly lower energy compared to that of mean field inference. We also show the effectiveness of our approach on semantic segmentation task. 

Speaker: Yuri Boykov

Title: Kernel Clustering meets Graphical Models

Abstract: This talk discusses two seemingly unrelated data analysis methodologies: kernel clustering and graphical models. Clustering is an unsupervised learning technique for general data where kernel methods are known for their discriminating power. Graphical models such as Markov Random Fields (MRF) and related continuous geometric methods represent common image segmentation methodologies. While both clustering and regularization models are very widely used in machine learning and computer vision, they could not be combined before due to significant differences in the corresponding optimization, e.g. spectral relaxation vs. combinatorial optimization methods. This talk reviews the general properties of kernel clustering and graphical models, discusses their limitations (including newly discovered "density biases" in kernel methods), and proposes a general easy-to-implement algorithm based on iterative bound optimization. In particular, we show that popular MRF potentials introduce principled geometric and contextual constraints into clustering, while standard kernel methodology allows graphical models to work with arbitrary high-dimensional features (e.g. RGBD, RGBDXY, deep, etc).