Accepted to AAAI 2023 [Paper]
Xiaohang Wang*, Xuanhong Chen*, Bingbing Ni**, Zhengyan Tong, Hang Wang
* Equal contribution
** Corresponding author
The official repository with Pytorch
- This work is for arbitrary-scale RGB-guided depth map super-resolution (DSR).
- Depth map super-resolution (DSR) has been a fundamental task for 3D computer vision. While arbitrary scale DSR is a more realistic setting in this scenario, previous approaches predominantly suffer from the issue of inefficient real-numbered scale upsampling.
- python3.7+
- pytorch1.9+
- torchvision
- Nvidia Apex (python-only build is ok.)
We follow Tang et al. and use the same datasets. Please refer to here to download the preprocessed datasets and extract them into data/
folder.
-
Baidu Netdisk (百度网盘):https://pan.baidu.com/s/1e2rLQFqVHIy2ZZG922XNTA
-
Extraction Code (提取码):xu7e
-
Google Drive: https://drive.google.com/drive/folders/1cIvA_AYh0fve_pDhN6timhCeN6A7MhD2?usp=share_link
Please put the model under workspace/checkpoints
folder.
python main.py
bash test.sh
For academic and non-commercial use only. The whole project is under the MIT license. See LICENSE for additional details.
If you find this project useful in your research, please consider citing:
@misc{GeoDSR,
author = {Wang, Xiaohang and Chen, Xuanhong and Ni, Bingbing and Tong, Zhengyan and Wang, Hang},
title = {Learning Continuous Depth Representation via Geometric Spatial Aggregator},
publisher = {arXiv},
year = {2022}
}
This code is built based on JIIF. We thank the authors for sharing the codes.
Learn about our other projects
[EQSR]: high-quality arbitrary-scale image super-resolution;
[VGGFace2-HQ]: high resolution face dataset VGGFace2-HQ;
[SimSwap]: most popular face swapping project;
[ASMA-GAN]: high-quality style transfer project;