ECCVW-2020 paper "Self-Attention based Feature Extractors for 3D Object Detection in Point Clouds"
Aug 18, 202021 views
We study the effectiveness of elf-attention based featurizers in the task of 3D object detection for point clouds. Our experiments reveal that self-attention based featurizers exhibit a promising performance compared to other featurizers (PointNet, PointNet++, StarNet, and so on) that rely on point-wise transformations. This paper is accepted in the Workshop on Perception for Autonomous Driving - ECCV 2020.