Delving into Dark Regions for Robust Shadow Detection

Keywords Shadow detection · Global/local shadow understanding · Region-based shadow analysis

This work was done in 2021.

[paper] [github]

Resources

Download model weights and put into folder "models"

SBU: https://drive.google.com/file/d/1TY8O5F9GPB0Zv7CxQoAUaqQTbJFLdb6a/view?usp=sharing

ISTD: https://drive.google.com/file/d/1tc6WpmlAwZTg7aJL18moz2bBn9Rxves0/view?usp=sharing

Demo

We provide a demo in jupyter notebook format "demo.ipynb". Start the jupyter-lab and run all cells for a quick start!

Inference

1. configurate "config.py"
2. run "infer.py"

BER Score:
SBU: 3.04
UCF: 7.75
ISTD: 1.33

If you want to train our model, please refer to other/train_script.rar for details (todo: make it better for revision).

Evaluation

1. After the inference, set the tool/config.py
2. Run "cd tool; python start.py"

Cite

If you find our work is helpful, cite and star it! Thanks!

@misc{guan2024delving,
      title={Delving into Dark Regions for Robust Shadow Detection}, 
      author={Huankang Guan and Ke Xu and Rynson W. H. Lau},
      year={2024},
      eprint={2402.13631},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}