LoDa (CVPR 2024)

This is the official repository of the paper "Boosting Image Quality Assessment through Efficient Transformer Adaptation with Local Feature Enhancement".

Updates

• [06/2024] We released the source code of 'LoDa', check the code on GitHub

To-Dos

• Checkpoints & Logs
• Initialization

Usage

Pre-requisition

Installation

We recommend using the conda package manager to avoid dependency problems.

1. Clone the repository

git clone https://github.com/NeosXu/LoDa

2. Install Python dependencies

# Using conda (Recommend)
conda env create -f environment.yaml
conda activate loda

# Using pip
pip install -r requirements.txt
pip install -r requirements-dev.txt # Optional, for code formatting

pre-commit install # Optional, for code formatting

Data Preparation

You need to download the corresponding datasets in the paper and place them under the same directory data.

For each dataset, run the corresponding preprocess script to process the image, metadata and train/test split of the datasets.

dataset_names=("live" "tid2013" "kadid10k" "livechallenge" "koniq10k" "spaq" "flive")
for dn in "${dataset_names[@]}"
do
    python scripts/process_"$dn".py
done

At the end, the directory structure should look like this:

├── data
|    ├── flive
|    ├── kadid10k
|    ├── koniq10k
|    ├── live_iqa
|    ├── LIVEC
|    ├── spaq
|    ├── tid2013
|    ├── meta_info
|    |    ├── meta_info_FLIVEDataset.csv
|    |    ├── meta_info_KADID10kDataset.csv
|    |    ├── meta_info_KonIQ10kDataset.csv
|    |    ├── ...
|    ├── train_split_info
|    |    ├── flive_82_seed3407.pkl
|    |    ├── kadid10k_82_seed3407.pkl
|    |    ├── koniq10k_82_seed3407.pkl
|    |    ├── ...

Or you can simply download the meta_info and train_split_info from Google Drive.

Training

mkdir logs
# all datasets
bash scripts/benchmark/benchmark_loda_all.sh 0
# a single dataset
bash scripts/benchmark/benchmark_loda_koniq10k.sh 0

Evaluation

mkdir logs
# all datasets
bash scripts/benchmark/benchmark_loda_eval_all.sh 0

Citing LoDa

If you find this project helpful in your research, please consider citing our papers:

@InProceedings{Xu_2024_CVPR,
    author    = {Xu, Kangmin and Liao, Liang and Xiao, Jing and Chen, Chaofeng and Wu, Haoning and Yan, Qiong and Lin, Weisi},
    title     = {Boosting Image Quality Assessment through Efficient Transformer Adaptation with Local Feature Enhancement},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    month     = {June},
    year      = {2024},
    pages     = {2662-2672}
}

Acknowledgement

We borrowed some parts from the following open-source projects:

• IQA-PyTorch
• pytorch-project-template

Many thanks to them.