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

Expand Torchgeo_FCN_Full.ipynb use more Chesapeake datasets #185

Open
Pale-Blue-Dot-97 opened this issue Jun 1, 2023 · 1 comment
Open
Assignees
Labels
enhancement New feature or request

Comments

@Pale-Blue-Dot-97
Copy link
Owner

Pale-Blue-Dot-97 commented Jun 1, 2023

Issue

The Torchgeo_FCN_Full.ipynb is supposed to be a demonstration of using a wide range of Chesapeake dataset but does not currently do so.

Solution

Expand Torchgeo_FCN_Full.ipynb to use more Chesapeake datasets.

@Pale-Blue-Dot-97 Pale-Blue-Dot-97 added the enhancement New feature or request label Jun 1, 2023
@Pale-Blue-Dot-97 Pale-Blue-Dot-97 self-assigned this Jun 1, 2023
@Pale-Blue-Dot-97 Pale-Blue-Dot-97 changed the title Expand Torchgeo_Demo_Full.ipynb use more Chesapeake datasets Expand Torchgeo_FCN_Full.ipynb use more Chesapeake datasets Jun 1, 2023
@Pale-Blue-Dot-97
Copy link
Owner Author

It appears that the ChesapeakeCVPR dataset class is still quite buggy. microsoft/torchgeo#619 highlights issues with the patch sizes misaligning during sampling. Supposedly, this is fixed in microsoft/torchgeo#992 by sampling a double the patch size then cropping down in ChesapeakeCVPRDataModule.

However, I get an error in the cropping:

AssertionError: Crop size must be smaller than input size. Got (198, 106) and (128,
128).

This occurs whether I set patch_size=128 or patch_size=256. I may lodge an issue on torchgeo to see if they can assist and fix this bug.

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

No branches or pull requests

1 participant