Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

[Feat] kornia.Resize has a bad default for align_corners

See original GitHub issue

🐛 Bug

kornia.Resize has a bad default for align_corners if interpolation=="nearest" or interpolation=="area".

To Reproduce

import torch
import kornia

transform = kornia.Resize(32, interpolation="nearest")

input = torch.rand(1, 3, 16, 16)
transform(input)

ValueError: align_corners option can only be set with the interpolating modes: linear | bilinear | bicubic | trilinear

Expected behavior

This should work out of the box. You need to explicitly set align_corners=None to get this working. This is neither documented nor even accepted by the type annotations.

Environment

kornia==0.4.1

Issue Analytics

State:
Created 3 years ago
Comments:9 (6 by maintainers)

Top GitHub Comments

1reaction

adamjstewartcommented, Nov 27, 2022

This appears to be fixed in the latest release. The following has no errors:

transform = kornia.geometry.transform.Resize(32, interpolation="nearest")

although the original error can be reproduced with:

transform = kornia.geometry.transform.Resize(32, interpolation="nearest", align_corners=True)

1reaction

pmeiercommented, Dec 1, 2020

Sorry for the delay. I can give it a shot.

Top Results From Across the Web

kornia.geometry.transform - Read the Docs

align_corners ( bool , optional) – interpolation flag. Default: True ... kornia.geometry.transform.scale(tensor, scale_factor, center=None, mode='bilinear', ...