Eduard Trulls is currently a post-doc at the Computer Vision Lab at EPFL in Lausanne, Switzerland. He obtained his PhD from the Institute of Robotics in Barcelona, Spain, in 2015. His thesis explored novel strategies to enhance local, low-level features (e.g. SIFT, HOG) with global, mid-level data such as motion or segmentation cues. His current work focuses on designing novel approaches to apply deep learning techniques to classical, low-level computer vision problems such as local feature extraction and matching for 3D reconstruction. In this talk, Eduart will present a novel deep architecture to learn to find good correspondences for wide-baseline stereo. Our solution is based on putative keypointmatches, which we learn to label as inliers or outliers while simultaneously using them to recover the camera pose, a fundamental Computer Vision problem. Our solution is simple (no convolutional or fully-connected layers), small (4 Mb), easy to train (state of the art matching outdoors scenes with only 59 images) and generalizes well, particularly in contrast to dense networks which require the entire image.

Zoom International Numbers: https://zoom.us/zoomconference

Check the website in case the date or time changes: http://cognitive-science.info/community/weekly-update/  & where you can find slides & recordings of prior calls)

Please retweet https://twitter.com/sumalaika/status/1036933287640489984

Join LinkedIn Group https://www.linkedin.com/groups/6729452