Sparse and Dense Data with CNNs: Depth Completion and Semantic Segmentation

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.
Document type :
Conference papers
Complete list of metadatas

https://hal.archives-ouvertes.fr/hal-01858241
Contributor : Maximilian Jaritz <>
Submitted on : Monday, August 20, 2018 - 11:54:38 AM
Last modification on : Tuesday, November 27, 2018 - 2:26:55 PM

Links full text

Identifiers

  • HAL Id : hal-01858241, version 1
  • ARXIV : 1808.00769

Collections

Citation

Maximilian Jaritz, Raoul de Charette, Emilie Wirbel, Xavier Perrotton, Fawzi Nashashibi. Sparse and Dense Data with CNNs: Depth Completion and Semantic Segmentation. 3DV 2018 - 6th international conference on 3D Vision, Sep 2018, Verona, Italy. ⟨hal-01858241⟩

Share

Metrics

Record views

112