Pytorch implementation for (PRCV 2023) ContextNet: Learning Context Information for Texture-less Light Field Depth Estimation.

Ubuntu            16.04
Python            3.8.10
Tensorflow-gpu    2.5.0
CUDA              11.2

1. Download UrbanLF-Syn dataset
2. Run python train_urban.py to train model

• Checkpoint files will be saved in 'LF_checkpoints/XXX_ckp/iterXXXX_valmseXXXX_bpXXX.hdf5'.
• Training process will be saved in
  • 'LF_output/XXX_ckp/train_iterXXXXX.jpg'
  • 'LF_output/XXX_ckp/val_iterXXXXX.jpg'.

3. Run python evaltion_urban.py

• path_weight='pretrained_contextnet.hdf5'

• Run python submission_urban.py
  • path_weight='pretrained_contextnet.hdf5'

The pretrained weights are available in the https://drive.google.com/file/d/1tRKyA74IzwETa4RLGxrSHFmYBzrpVU6q/view?usp=drive_link.

If you find this work helpful, please consider citing:

@inproceedings{chao2023contextnet,
  title={ContextNet: Learning Context Information for Texture-less Light Field Depth Estimation},
  author={Chao, Wentao and  Wang, Xuechun and Kan, Yiming and Duan, Fuqing},
  booktitle={Chinese Conference on Pattern Recognition and Computer Vision (PRCV)},
  year={2023},
  organization={Springer}
}

This code borrows heavily from SubFocal repository. Thanks a lot.