TSHRNet - MindSpore

MindSpore implement of

Towards High-Quality Specular Highlight Removal by Leveraging Large-scale Synthetic Data

Gang Fu, Qing Zhang, Lei Zhu, Chunxia Xiao, and Ping Li

In ICCV's 23

Paper

In this paper, our goal is to remove specular highlight removal for object-level images. In this paper, we propose a three-stage network for specular highlight removal, consisting of (i) physics-based specular highlight removal, (ii) specular-free refinement, and (iii) tone correction. In addition, we present a large-scale synthetic dataset of object-level images, in which each input image has corresponding albedo, shading, specular residue, diffuse, and tone-corrected diffuse images.

Prerequisities of MindSpore implementation

mindspore = 2.1.1
matplotlib
pillow
tqdm

Datasets

• Our SHHR dataset is available at OneDrive or Google Drive (~5G).

Pretrained models

• Pretrained MindSpore models on SSHR are available at checkpoints_SSHR_ms.
• Also available at release.

Training

python train.py \
       -trdd ${train_data_dir} \
       -trdlf ${train_data_list_file} \
       -dn ${dataset_name}

Testing

Note thatwe split "test.lst" into four parts for testin, due to out of memory.

python test.py \
       -mn ${model_name} \
       -l ${num_checkpoint} \
       -tdn ${testing_data_name} \
       -tedd ${testing_data_dir} \
       -tedlf ${testing_data_list_file}

Citation

@inproceedings{fu-2023-towar-a,
  author =     {Fu, Gang and Zhang, Qing and Zhu, Lei and Xiao, Chunxia and Li, Ping},
  title =     {Towards high-quality specular highlight removal by leveraging large-scale synthetic data},
  booktitle =     {Proceedings of the IEEE International Conference on Computer Vision},
  year =     {2023},
  pages =     {To appear},
}