eduzhai > Applied Sciences > Engineering >

Associate-3Ddet Perceptual-to-Conceptual Association for 3D Point Cloud Object Detection

  • Save

... pages left unread,continue reading

Document pages: 10 pages

Abstract: Object detection from 3D point clouds remains a challenging task, thoughrecent studies pushed the envelope with the deep learning techniques. Owing tothe severe spatial occlusion and inherent variance of point density with thedistance to sensors, appearance of a same object varies a lot in point clouddata. Designing robust feature representation against such appearance changesis hence the key issue in a 3D object detection method. In this paper, weinnovatively propose a domain adaptation like approach to enhance therobustness of the feature representation. More specifically, we bridge the gapbetween the perceptual domain where the feature comes from a real scene and theconceptual domain where the feature is extracted from an augmented sceneconsisting of non-occlusion point cloud rich of detailed information. Thisdomain adaptation approach mimics the functionality of the human brain whenproceeding object perception. Extensive experiments demonstrate that our simpleyet effective approach fundamentally boosts the performance of 3D point cloudobject detection and achieves the state-of-the-art results.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...