Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

about data process #47

Open
DIY-Z opened this issue Aug 17, 2023 · 0 comments
Open

about data process #47

DIY-Z opened this issue Aug 17, 2023 · 0 comments

Comments

@DIY-Z
Copy link

DIY-Z commented Aug 17, 2023

hyperIQA/data_loader.py

Lines 31 to 35 in 685d4af

elif dataset == 'koniq-10k':
if istrain:
transforms = torchvision.transforms.Compose([
torchvision.transforms.RandomHorizontalFlip(),
torchvision.transforms.Resize((512, 384)),

sample.append((os.path.join(root, '1024x768', imgname[item]), mos_all[item]))

From the two code snippets above, it is evident that the data is loaded as images with a size of '1024x768', with width and height dimensions of 1024 and 768 respectively. However, the Resize((512, 384)) operation rescales the dimensions to 512 and 384, resulting in a noticeable change in the aspect ratio from the original 768:1024 to 512:384. I'm curious if the same processing is applied in the experimental setup of the paper?

Plus: according to the document of pytorch Resize,the 'size' parameter of the Resize function refers to the height and width.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant