Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Tangent Convolutions for Dense Prediction in 3D

About

We present an approach to semantic scene analysis using deep convolutional networks. Our approach is based on tangent convolutions - a new construction for convolutional networks on 3D data. In contrast to volumetric approaches, our method operates directly on surface geometry. Crucially, the construction is applicable to unstructured point clouds and other noisy real-world data. We show that tangent convolutions can be evaluated efficiently on large-scale point clouds with millions of points. Using tangent convolutions, we design a deep fully-convolutional network for semantic segmentation of 3D point clouds, and apply it to challenging real-world datasets of indoor and outdoor 3D environments. Experimental results show that the presented approach outperforms other recent deep network constructions in detailed analysis of large 3D scenes.

Maxim Tatarchenko, Jaesik Park, Vladlen Koltun, Qian-Yi Zhou• 2018

Related benchmarks

TaskDatasetResultRank
Semantic segmentationS3DIS (Area 5)
mIOU52.8
907
Semantic segmentationS3DIS (6-fold)
mIoU (Mean IoU)52.8
344
Semantic segmentationSemanticKITTI (test)
mIoU40.9
335
Semantic segmentationScanNet (val)
mIoU49
274
3D Semantic SegmentationScanNet V2 (val)
mIoU40.9
209
3D Semantic SegmentationScanNet v2 (test)
mIoU43.8
110
3D Semantic SegmentationScanNet (test)
mIoU43.8
109
3D Semantic SegmentationScanNet v1 (test)--
72
Semantic segmentationScanNet (test)
mIoU43.8
64
LiDAR Semantic SegmentationSemanticKITTI 1.0 (test)
mIoU40.9
59
Showing 10 of 28 rows

Other info

Code

Follow for update