search

megvii-model / HINet

Other
367 stars 62 forks source link

wrong display: Validation SIDD_val, # psnr: 23.2228 #41

Open zxcvads opened 2 years ago

zxcvads commented 2 years ago

hi, author ,thanks for your enjoy, it's a nice work! But when i do reproduce of HINet in sidd dataset, during training(train.py) i can get psnr39.9124, in(test.py)i used the trained model, got the wrong display: Validation SIDD_val, # psnr: 23.2228, i don't know what went wrong?

mayorx commented 2 years ago

Hi, nono0822,

Thanks for your attention to HINet. Could you please provide more details e.g. test logs/ your test configs? And please check whether the model parameters have been properly loaded. Thanks!

zxcvads commented 2 years ago

update: I use the pretrain model to test, "python basicsr/test.py -opt options/test/SIDD/HINet-SIDD-1x.yml" but got the wrong psnr, i'm confused.... 2022-07-18 17:21:31,459 INFO: name: HINet-SIDD model_type: ImageRestorationModel scale: 1 num_gpu: 1 manual_seed: 10 datasets:[ test:[ name: SIDD_val type: PairedImageDataset_SIDD dataroot_gt: ./datasets/SIDD/val/gt_crops.lmdb dataroot_lq: ./datasets/SIDD/val/input_crops.lmdb io_backend:[ type: lmdb ] phase: test scale: 1 ] ] network_g:[ type: HINet wf: 64 hin_position_left: 0 hin_position_right: 4 ] path:[ pretrain_network_g: ./experiments/pretrained_models/HINet-SIDD-1x.pth strict_load_g: True resume_state: None root: ./HINet-main results_root: ./HINet-main/results/HINet-SIDD log: ./HINet-main/results/HINet-SIDD visualization:./HINet-main/results/HINet-SIDD/visualization ] val:[ save_img: False rgb2bgr: True use_image: False metrics:[ psnr:[ type: calculate_psnr crop_border: 0 test_y_channel: False ] ssim:[ type: calculate_ssim crop_border: 0 test_y_channel: False ] ] ] dist_params:[ backend: nccl port: 29500 ] is_train: False dist: False rank: 0 world_size: 1

2022-07-18 17:21:31,469 INFO: Dataset PairedImageDataset_SIDD - SIDD_val is created. 2022-07-18 17:21:31,469 INFO: Number of test images in SIDD_val: 1280 2022-07-18 17:21:41,059 INFO: Network: HINet, with parameters: 88,669,702 ...... 2022-07-18 17:21:41,060 INFO: Model [ImageRestorationModel] is created. 2022-07-18 17:21:41,061 INFO: Testing SIDD_val... 2022-07-18 17:30:56,869 INFO: Validation SIDD_val, # psnr: 21.1465 # ssim: 0.4399