Data and codes for "SCANet: Self-paced semi-curricular attention network for non-homogeneous image dehazing"

This record contains the data and codes for the paper "SCANet: Self-Paced Semi-Curricular Attention Network for Non-Homogeneous Image Dehazing" published in <i>2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)</i>. RequirementPython 3.7Pytorch 1.9.1Network ArchitectureTrainPlace the training and test image pairs in the <code>data</code> folder.Run <code>data/makedataset.py</code> to generate the <code>NH-Haze20-21-23.h5</code> file.Run <code>train.py</code> to start training.TestPlace the pre-training weight in the <code>checkpoint</code> folder.Place test hazy images in the <code>input</code> folder.Modify the weight name in the <code>test.py</code>.<br><pre><pre>parser.add_argument("--model_name", type=str, default='Gmodel_40', help='model name')<br></pre></pre>Run <code>test.py</code>The results is saved in <code>output</code> folder.Pre-training Weight DownloadThe weight40 <code>Gmodel_40.tar</code> for the NTIRE2023 val/test datasets, i.e., the weight used in the NTIRE2023 challenge.The weight105 <code>Gmodel_105.tar</code> for the NTIRE2020/2021/2023 datasets.The weight120 <code>Gmodel_120.tar</code> for the NTIRE2020/2021/2023 datasets (Add the 15 tested images as the training dataset).