SPLINE-Net: Sparse Photometric Stereo through Lighting Interpolation and Normal Estimation Networks
http://openaccess.thecvf.com/content_ICCV_2019/papers/Zheng_SPLINE-Net_Sparse_Photometric_Stereo_Through_Lighting_Interpolation_and_Normal_Estimation_ICCV_2019_paper.pdf
- Python 3.5+
- PyTorch 0.4.0+
- TensorFlow 1.3+
- See solver.py
Please refer to generate_data_in_one.m.
# Prepare the test set which consists of 100 subsets, 10 lightings each object
The data can be downloaded from: https://pan.baidu.com/s/1UQVBLpwnzn-Fn76QwHDbgA, code：iccv
# Download pre-trained model
The pre-trained model can be downloaded from: https://pan.baidu.com/s/1OiAc76HgZA9s4NzYxxYRoA, code: iccv
# Run for test
python main.py --mode test
# Please check the results in photometric/results
- Please follow the data format of test set we created, use 'data/test' as a reference.
- In text file 'data/test/.../..txt', each line is a 1 * 24 vector, elem1, ..., elem24, represent data of one pixel.
- Elem1 is index of the pixel in the original image. Elem2, elem3, elem4 are nx, ny, nz in normal vector.
- Elem5, ..., elem14 are index of the observation map, mapping to 10 lightings.
- Elem15, ..., elem24 are corresponding intensities.
If you find our code is useful, please cite our paper. If you have any problem of implementation or running the code, please contact us: csqianzheng@gmail.com, jiaym15@outlook.com
@inproceedings{zheng2019spline,
title={SPLINE-Net: Sparse photometric stereo through lighting interpolation and normal estimation networks},
author={Zheng, Qian and Jia, Yiming and Shi, Boxin and Jiang, Xudong and Duan, Ling-Yu and Kot, Alex C},
booktitle={Proceedings of the IEEE International Conference on Computer Vision},
pages={8549--8558},
year={2019}
}