From Sky to the Ground: A Large-scale Benchmark and Simple Baseline Towards Real Rain Removal

ICCV 2023

Yun Guo†

Xueyao Xiao†

Yi Chang*

Shumin Deng

Luxin Yan

School of Artificial Intelligence and Automation, Huazhong University of Science and Technology, China

[Paper]

[Github]

[Supplementary]

Benchmark Examples

LHP-Rain dataset contains various rain patterns, such as streaks, veiling, occlusion, and the first dataset to claim and tackle ground splashing problem.

Benchmark Statistics

(a) Distribution of rain and scene of the proposed benchmark. (b) Our proposed LHP-Rain outperforms others in terms of rain diversity and sequence amount.

Benchmark Collection

I. Capture real rain videos with static background.
II. Pick out moving object to remove unexpected disturbance.
III. Employ the proposed RLRTR to obtain high-quality GT.

Demo

Section 1: Examples of our collected static rainy videos and GT generated by RLRTR. Section 2: Performance of SCD-Former on real-world rainy images. Please change the video quality to 1080p for better visualization.

Update

• 2024.04.07 Repair some issues in the dataset.

Download

• Full size (3000 rainy/clean image pairs with 1920*1080 resolution. Note that few images are cropped smaller than original size.)
Baidu NetDisk (code: jg58) / Google Drive
• Patch size: Complete version (Complete train/val with 1 million 256*256 rainy/clean image pairs)
Baidu NetDisk (code: p61z) / Google Drive
• Patch size: Simple version (10% of original 256*256 train/val for effecient training)
Baidu NetDisk (code: 01gc) / Google Drive
• Patch size: Test set (1000 512*512 rainy/clean image patches for evaluating)
Baidu NetDisk (code: zwbj) / Google Drive
• High-level annotations (object detection/lane segmentation)
To be continued.

Acknowledgment

Code for this platform borrows from NeRF-W project. Thanks to Xiaoxiong Wang (HUST) for supporting this project.

Citation

@InProceedings{Guo_2023_ICCV, author = {Guo, Yun and Xiao, Xueyao and Chang, Yi and Deng, Shumin and Yan, Luxin}, title = {From Sky to the Ground: A Large-scale Benchmark and Simple Baseline Towards Real Rain Removal}, booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)}, month = {October}, year = {2023}, pages = {12097-12107} }