Skip to content

Latest commit

 

History

History
94 lines (69 loc) · 2.64 KB

README.md

File metadata and controls

94 lines (69 loc) · 2.64 KB
Raw

Joint Upsampling for Refocusing Light Fields Derived With Hybrid Lenses

The official implementation of paper "Joint Upsampling for Refocusing Light Fields Derived With Hybrid Lenses". You can visit our paper in link

Environment introduction

Hardware

CPU: Intel(R) Core(TM) i7-8700K
GPU: GTX 1080
RAM: 8G*2 2666

Python

torchvision==0.7.0+cu101
torch==1.6.0+cu101
opencv_python==4.4.0.46
numpy==1.18.5
scikit_image==0.18.1
Pillow==8.1.1
skimage==0.0

Dataset

The raw Lytro Dataset can be downloaded from Google Drive. You can use Lytro Desktop to refocus the raw lytro photos. Our dataset contains scenes with thin structures and rich textures (see below), which are difficult for refocused image upsampling.

Baidu, code: 4ca5 Google

Input (GT) in Training Dataset Guidance (GT) in Training Dataset
Input (GT) in Testing Dataset Guidance (GT) in Testing Dataset
Input (GT) in Additional Dataset Guidance (GT) in Additional Dataset

Model

The pretrained model is already uploaded in repo, ./Model/LFN.pth

Training

Customize the trainConfig in train.py and run it

python train.py

Testing

Customize the evalPngConfig in test.py and run it

python test.py

In test.py, you can use testAllInOne to test x2/x4/x8 for both shallow & deep testing dataset at one time. Or use evalPng to test the selected scale.

Other

If you have any question, please leave an issue.

Citation

@ARTICLE{10064040,
  author={Yang, Yang and Wu, Lianxiong and Zeng, Lanling and Yan, Tao and Zhan, Yongzhao},
  journal={IEEE Transactions on Instrumentation and Measurement}, 
  title={Joint Upsampling for Refocusing Light Fields Derived With Hybrid Lenses}, 
  year={2023},
  volume={72},
  number={},
  pages={1-12},
  doi={10.1109/TIM.2023.3253880}}