This is the repository of the WSRD dataset, used as benchmark for the NTIRE2023 Challenge on Image Shadow Removal.

The collection of 1000 roughly aligned image pairs can be found using the following URLS: train input link | train gt link.

We provide a database of 100 images, that were used for evaluation on the provided Codalab server: valid input link | valid gt link.

The input images used for testing in the Final Phase of the challenge are available here. Since we are planning for a second edition of the challenge, the test ground-truth images will remain private, for the moment. Follow the current repository for more updates.

Along with the provided data, we also provide results for DNSR, a baseline optimized for reconstruction fidelity.

• Checkpoints for DNSR are available here.
• Results of the DNSR over the WSRD benchmark are available here.
• Additional results for the ISTD/ISTD+ benchmarks are available here.

Please follow the current repository for more updates.

A version of the WSRD Dataset will be used as a benchmark for the NTIRE24 Challenge on Image Shadow Removal. The challenge has a fidelity track and a perceptual track.

This new version proposes improved pixel-alignment through homography estimation. train input link | train gt link.

The validation split is used in the Development Phase of the challenge. Here, you can download the input images | ground truth images.

The test split will be used in the final Test Phase. Since we are aiming for proper evaluation on unseen data, these images will stay, for the moment, private.

To test your model on both validation and testing splits, you can use the Codalab competition which will remain open. Results comparing the teams participating in the 2024 challenge are also available.

For access and other requests feel free to drop us an email.

Copyright (c) 2025 Computer Vision Lab, University of Wurzburg

Licensed under CC BY-NC-SA 4.0 (Attribution-NonCommercial-ShareAlike 4.0 International) (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at

https://creativecommons.org/licenses/by-nc-sa/4.0/legalcode

The code is released for academic research use only. For commercial use, please contact Computer Vision Lab, University of Wurzburg. Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.