End-to-End Learning of Deep Structured Models for Semantic Segmentation

Abstract: The task of semantic segmentation aims at understanding an image at a pixel level. This means assigning a label to each pixel of an image, describing the object it is depicting. Due to its applicability in many areas, such as autonomous vehicles, robotics and medical surgery assistance, semantic segmentation has become an essential task in image analysis. During the last few years a lot of progress have been made for image segmentation algorithms, mainly due to the introduction of deep learning methods, in particular the use of Convolutional Neural Networks (CNNs). CNNs are powerful for modeling complex connections between input and output data but lack the ability to directly model dependent output structures, for instance, enforcing properties such as label smoothness and coherence. This drawback motivates the use of Conditional Random Fields (CRFs), widely applied as a post-processing step in semantic segmentation. This thesis summarizes the content of three papers, all of them presenting solutions to semantic segmentation problems. The applications have varied widely and several different types of data have been considered, ranging from 3D CT images to RGB images of horses. The main focus has been on developing robust and accurate models to solve these problems. The models consist of a CNN capable of learning complex image features coupled with a CRF capable of learning dependencies between output variables. Emphasis has been on creating models that are possible to train end-to-end, as well as developing corresponding optimization methods needed to enable efficient training. End-to-end training gives the CNN and the CRF a chance to learn how to interact and exploit complementary information to achieve better performance.