search

jinyeying / DC-ShadowNet-Hard-and-Soft-Shadow-Removal

[ICCV2021]"DC-ShadowNet: Single-Image Hard and Soft Shadow Removal Using Unsupervised Domain-Classifier Guided Network", https://arxiv.org/abs/2207.10434
212 stars 21 forks source link

How to keep the processed images not resized? #21

Closed Alive59 closed 1 year ago

Alive59 commented 1 year ago

Thanks for the great code. I'm just wondering how to make the outputs the same dimension as the input images. At this stage the model would resize all inputs into 256*256.

jinyeying commented 1 year ago

Thanks for the great code. I'm just wondering how to make the outputs the same dimension as the input images. At this stage the model would resize all inputs into 256*256.

python main_test.py --dataset SRD --datasetpath [path_to_SRD dataset] --use_original_name True --im_suf_A .jpg

Alive59 commented 1 year ago

Thanks for the great code. I'm just wondering how to make the outputs the same dimension as the input images. At this stage the model would resize all inputs into 256*256.

python main_test.py --dataset SRD --datasetpath [path_to_SRD dataset] --use_original_name True --im_suf_A .jpg

Thanks for the reply, but it seems the model still works the same... Maybe I haven't described the question well but what I meant was that how to keep the output in the same dimension, for example 640 480, but not all in 256 256.

jinyeying commented 1 year ago

Thanks for the great code. I'm just wondering how to make the outputs the same dimension as the input images. At this stage the model would resize all inputs into 256*256.

python main_test.py --dataset SRD --datasetpath [path_to_SRD dataset] --use_original_name True --im_suf_A .jpg

Thanks for the reply, but it seems the model still works the same... Maybe I haven't described the question well but what I meant was that how to keep the output in the same dimension, for example 640 480, but not all in 256 256.

parser.add_argument('--img_h', type=int, default=480, help='The org size of image') parser.add_argument('--img_w', type=int, default=640, help='The org size of image') in [main_test.py]

self.img_size = args.img_size self.img_h = args.img_h self.img_w = args.img_w in [DCShadowNet_test.py]

Alive59 commented 1 year ago

I just found out that making the original images into smaller tiles leads to better results, so that's no longer a problem... Anyway, thanks for the help!