Attention-guided Non-Local Network (ATNLC) for High Dynamic Range Image Synthesis

Solution of HDR Challenge

ATNLC - NTIRE 2022 - Paper - tobeupdated...

by Tu Vo

Content

• ATNLC
• Getting Started
• Running
• References
• Citations

Getting Started

• Clone the repository

Prerequisites

• Tensorflow 2.7.0+
• Tensorflow_addons
• Python 3.7+
• Keras 2.3.0
• PIL
• numpy

Running

Training

• Preprocess

  • Download the training data

  • Unzip the file

  • Prepare data:

  python src/create_dataset_img.py
  

• Train ATNLC

  python main.py
  

Testing

• Test ATNLC

  python test.py
  

• Submit

  python test_submit.py
  

Calculate FLOPS and run-time

python calculate_flops.py

Usage

Training

usage: main.py  [-h] [--filter FILTER] 
                [--decoder_kernel DECODER_KERNEL]
                [--attention_filter ATTENTION_FILTER] 
                [--triple_pass_filter TRIPLE_PASS_FILTER]
                [--kernel KERNEL] [--encoder_kernel ENCODER_KERNEL] 

optional arguments:
  -h, --help            show this help message and exit
  --filter FILTER
  --attention_filter ATTENTION_FILTER
  --kernel KERNEL
  --encoder_kernel ENCODER_KERNEL
  --decoder_kernel DECODER_KERNEL
  --triple_pass_filter TRIPLE_PASS_FILTER

Testing

• Download the weight here and put it to the folder weights

See results

usage: test.py  [-h] [--filter FILTER] 
                [--decoder_kernel DECODER_KERNEL]
                [--attention_filter ATTENTION_FILTER] 
                [--triple_pass_filter TRIPLE_PASS_FILTER]
                [--kernel KERNEL] [--encoder_kernel ENCODER_KERNEL] 

optional arguments:
  -h, --help            show this help message and exit
  --test_path TEST_PATH
  --save_path SAVE_PATH
  --filter FILTER
  --gpu GPU
  --attention_filter ATTENTION_FILTER
  --kernel KERNEL
  --encoder_kernel ENCODER_KERNEL
  --decoder_kernel DECODER_KERNEL
  --triple_pass_filter TRIPLE_PASS_FILTER

Result:

Short image	Medium Image	Long Image	Output

• path to the full results: GGDrive

Submit:

usage: test_submit.py  [-h] [--filter FILTER] 
                [--decoder_kernel DECODER_KERNEL]
                [--triple_pass_filter TRIPLE_PASS_FILTER]
                [--kernel KERNEL] [--encoder_kernel ENCODER_KERNEL] 
                [--attention_filter ATTENTION_FILTER] [--zip_name ZIP_NAME]

optional arguments:
  -h, --help            show this help message and exit
  --test_path TEST_PATH
  --save_path SAVE_PATH
  --filter FILTER
  --gpu GPU
  --attention_filter ATTENTION_FILTER
  --kernel KERNEL
  --encoder_kernel ENCODER_KERNEL
  --decoder_kernel DECODER_KERNEL
  --triple_pass_filter TRIPLE_PASS_FILTER
  --zip_name ZIP_NAME

License

This project is licensed under the MIT License - see the LICENSE file for details

References

[1] High Dynamic Range Image Synthesis Challenge - NTIRE2022

Citation: to be updated ...

Acknowledgments

• This work is heavily based on our previous work on Defocus Deblurred Challenge 2021.