C2ANet: Cross-scale and Cross-modality Aggregation Network for Scene Depth Super-Resolution

This repo implements the testing of depth upsampling networks for "C2ANet: Cross-scale and Cross-modality Aggregation Network for Scene Depth Super-Resolutionn" by Xinchen Ye and et al. at DLUT.

Installation

The code requires pytorch>= 1.7.0 and python=3.7.

The mmcv is request, which can be installed by

pip install mmcv-full -f https://download.openmmlab.com/mmcv/dist/{cu_version}/{torch_version}/index.html.

Please replace the original deform_conv.py with the file we provided.

quick test

python test_middlebury.py

download pretrained models and test data

Link of weights: https://pan.baidu.com/s/1145CqN68BaCyVTO6rczgTg pwd：3cad

Link of middleury test data：https://pan.baidu.com/s/1u09_U3ljVFjxwOvnboYpEw pwd：u0r3