BpCNet

PyTorch implementation of AAAI 2023 paper: "Take Your Model Further: A General Post-refinement Network for Light Field Disparity Estimation via BadPix Correction".

Requirements

python 3.6
pytorch 1.8.0
ubuntu 18.04

Installation

First you have to make sure that you have all dependencies in place.

You can create an anaconda environment called BpCNet using

conda env create -f BpCNet.yaml
conda activate BpCNet

Demo

Here, we train BpCNet on OACC-Net and perform refinement as a demo.

Dataset:

Light Field Dataset: We use HCI 4D Light Field Dataset for training and test. Please first download light field datasets, and put them into corresponding folders in data/HCInew.

Initial disparity map: We perform BpCNet on other LF disparity methods for refinement, you can provide initial data and put them into data/CoarseData or use the demo we provided.

To train, run:

python train.py --config configs/HCInew/BpCNet.yaml

To generate, run:

python generate.py --config configs/pretrained/HCInew/BpCNet_pretrained.yaml

If you find our code or paper useful, please consider citing:

waiting

crs904620522/BpCNet