sj-li/CSPN_monodepth

PyTorch implementation of Convolutional Spatial Propagation Network

CSPN implemented in Pytorch 0.4.1

Introduction

This is a PyTorch(0.4.1) implementation of Depth Estimation via Affinity Learned with Convolutional Spatial Propagation Network. At present, we can provide train script in NYU Depth V2 dataset for depth completion and monocular depth estimation. KITTI will be available soon!

Note: we fix some bugs in original code.

Reselt

We use 2 Titan X to train CSPN for depth completion and monocular depth estimation.

Monocular Depth Estimation

Method	rml	rmse	log10	Delta1	Delta2	Delta3
CSPN_ours	0.151	0.546	0.064	0.793	0.949	0.985

Depth Completion

We test CSPN for depth completion in NYU Depth dataset and use 500 sparse samples.

Method	rml	rmse	log10	Delta1	Delta2	Delta3
CSPN	0.016	0.117	-	0.992	0.999	1.000
CSPN_ours	0.023	0.152	0.010	0.988	0.997	0.999