Print Email Facebook Twitter Tilting at windmills Title Tilting at windmills: Data augmentation for deep pose estimation does not help with occlusions Author Pytel, Rafal (TU Delft Electrical Engineering, Mathematics and Computer Science) Contributor van Gemert, J.C. (mentor) Kayhan, O.S. (mentor) Reinders, M.J.T. (graduation committee) Liem, C.C.S. (graduation committee) Degree granting institution Delft University of Technology Date 2020-08-31 Abstract Occlusion degrades the performance of human pose estimation. In this paper, we introduce targeted keypoint and body part occlusion attacks. The effects of the attacks are systematically analyzed on the best-performing methods. In addition, we propose occlusion specific data augmentation techniques against keypoint and part attacks. Our extensive experiments show that human pose estimation methods are not robust to occlusion and data augmentation does not solve the occlusion problems. Subject Deep LearningComputer VisionHuman Pose EstimationOcclusionsData augmentations To reference this document use: http://resolver.tudelft.nl/uuid:cf50ef4b-801b-4645-aab5-31c55abf07a1 Embargo date 2020-08-24 Part of collection Student theses Document type master thesis Rights © 2020 Rafal Pytel Files PDF MSc_Thesis_Final.pdf 22.04 MB Close viewer /islandora/object/uuid:cf50ef4b-801b-4645-aab5-31c55abf07a1/datastream/OBJ/view