![]() ![]() With a piecewise bilinear mapping that is invertible. ToĮnable efficient and differentiable unzooming, we approximate the zooming warp In this work (LZU), we "learn to zoom" in on the input image,Ĭompute spatial features, and then "unzoom" to revert any deformations. However, for tasks with spatial labels (suchĪs 2D/3D object detection and semantic segmentation), such distortions may harm That "learn to zoom" on salient image regions, reducing compute while retaining Previous works propose nonuniform downsamplers Download a PDF of the paper titled Learning to Zoom and Unzoom, by Chittesh Thavamani and 3 other authors Download PDF Abstract: Many perception systems in mobile computing, autonomous navigation, and AR/VRįace strict compute constraints that are particularly challenging for
0 Comments
Leave a Reply. |