ISPDiffuser

This is the official employment of "ISPDiffuser: Learning RAW-to-sRGB Mappings with Texture-Aware Diffusion Models and Histogram-Guided Color Consistency"

Inferece

1. Download the checkpoint of ZRR dataset and MAI dataset from Baidu Netdisk or GoogleDrive, and put them into ckpt floder.

2. Download ZRR dataset and MAI dataset. MAI dataset can be download in Baidu Netdisk or GoogleDrive.

3. Change the dataset dir in configs/MAI_dataset.yml and configs/ZRR_dataset.yml

4. Run python evaluate.py --config path_to_ISPDiffuser/configs/MAI_dataset.yml

資料格式讀取 txt檔案 requiment檔案 patch_size 設定問題 win_size 問題 ssim_gt = structural_similarity(recon_img, y, win_size=7, data_range=255, multichannel=True, gaussian_weights=True) ISPDiffuser img_id 問題

config 缺乏scale Traceback (most recent call last): File "D:\github_code\ISPDiffuser-main\train.py", line 74, in main() File "D:\github_code\ISPDiffuser-main\train.py", line 70, in main diffusion.train(DATASET) File "D:\github_code\ISPDiffuser-main\models\ISPDiffuser.py", line 238, in train filename=os.path.join(self.config.data.ckpt_dir, str(self.config.data.scale), 'model_latest')) AttributeError: 'Namespace' object has no attribute 'scale'