Sparse and Dense Data with CNNs: Depth Completion and Semantic Segmentation
Computer Vision and Pattern Recognition
2018-09-03 v2
Abstract
Convolutional neural networks are designed for dense data, but vision data is often sparse (stereo depth, point clouds, pen stroke, etc.). We present a method to handle sparse depth data with optional dense RGB, and accomplish depth completion and semantic segmentation changing only the last layer. Our proposal efficiently learns sparse features without the need of an additional validity mask. We show how to ensure network robustness to varying input sparsities. Our method even works with densities as low as 0.8% (8 layer lidar), and outperforms all published state-of-the-art on the Kitti depth completion benchmark.
Cite
@article{arxiv.1808.00769,
title = {Sparse and Dense Data with CNNs: Depth Completion and Semantic Segmentation},
author = {Maximilian Jaritz and Raoul de Charette and Emilie Wirbel and Xavier Perrotton and Fawzi Nashashibi},
journal= {arXiv preprint arXiv:1808.00769},
year = {2018}
}
Comments
3DV 2018