CNNIQA

PyTorch 1.3 implementation of the following paper: Kang L, Ye P, Li Y, et al. Convolutional neural networks for no-reference image quality assessment[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2014: 1733-1740.

Note

• The optimizer is chosen as Adam here, instead of the SGD with momentum in the paper.
• the mat files in data/ are the information extracted from the datasets and the index information about the train/val/test split. The subjective scores of LIVE is from the realigned data.

Training

CUDA_VISIBLE_DEVICES=0 python main.py --exp_id=0 --database=LIVE

Before training, the im_dir in config.yaml must to be specified. Train/Val/Test split ratio in intra-database experiments can be set in config.yaml (default is 0.6/0.2/0.2).

Evaluation

Test Demo

python test_demo.py --im_path=data/I03_01_1.bmp

Cross Dataset

python test_cross_dataset.py --help

TODO: add metrics calculation. SROCC, KROCC can be easily get. PLCC, RMSE, MAE, OR should be calculated after a non-linear fitting since the quality score ranges are not the same across different IQA datasets.

Visualization

tensorboard --logdir=tensorboard_logs --port=6006 # in the server (host:port)
ssh -p port -L 6006:localhost:6006 user@host # in your PC. See the visualization in your PC

Requirements

conda create -n reproducibleresearch pip python=3.6
source activate reproducibleresearch
pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple
source deactive

• Python 3.6.8
• PyTorch 1.3.0
• TensorboardX 1.9, TensorFlow 2.0.0
• pytorch/ignite 0.2.1

Note: You need to install the right CUDA version.

TODO (If I have free time)

• Simplify the code
• Report results on some common databases
• etc.