Our team members have presented the following paper at the Computer Vision and Pattern Recognition Conference 2019 which took place from 16th to 20th of June 2019 in Long Beach, California, USA and is the “premier” conference in this field (Computer Vision).
SDC – Stacked Dilated Convolution: A Unified Descriptor Network for Dense Matching Tasks
René Schuster, Oliver Wasenmüller, Christian Unger, Didier Stricker
Content: To introduce a new design element for Deep Neural Networks (SDC – Stacked Dilated Convolutions) and beyond that to apply this element in a network for Dense Feature Description. With the new descriptor, we were able to improve matching for stereo disparity, optical flow, and scene flow on different data sets by up to 50%.
Our paper was accepted from 5160 submissions as one of 1294 (acceptance rate: 25.1 %) and was also one of the 288 oral papers (oral presentation in addition to the poster session), oral rate: 5.6 % of submissions, 22.3 % of accepted papers. The presentation can be seen here.