CNN Injected Transformer for Exposure Correction

Introduction

This repository is the official implementation of the paper, "CNN Injected Transformer for Exposure Correction".

Environment

• basicsr==1.4.2
• scikit-image==0.15.0

Dataset Preparation

• Download datasets

1. MSEC dataset (please refer to https://github.com/mahmoudnafifi/Exposure_Correction)

2. SICE dataset (please refer to https://github.com/KevinJ-Huang/ExposureNorm-Compensation)

• Extract image patches

python scripts/extract_subimages_MSEC.py

• Generate meta information

python scripts/generate_meta_info_MSEC.py

How to Test

• Download the pre-trained model

• Example: Testing on the MSEC dataset with images retouched by expert-a as ground truth

PYTHONPATH="./:${PYTHONPATH}" CUDA_VISIBLE_DEVICES=0 python test.py -opt options/test/Test_EC_MSEC_pretrained_over_expert_a_mlr.yml

How to train

• Single GPU training

PYTHONPATH="./:${PYTHONPATH}" CUDA_VISIBLE_DEVICES=0 python train.py -opt options/train/Train_EC_MSEC_win8_B8G1_30k_mlr.yml

• Distributed training

PYTHONPATH="./:${PYTHONPATH}" \
CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 \
python -m torch.distributed.launch --nproc_per_node=8 --master_port=4321 train.py -opt options/train/Train_EC_MSEC_win8_B32G4_30k_mlr.yml --launcher pytorch