Does XFeat work for two similar images but one of them is upsidedown?

[AI-Passionner]

I just found if two similar images (or same), if one of them is upside down, the matching points are horrible. It didn't work as SIFT, which matches the features between both images and can rotate back when aligning two images.

Am I wrong here? or any suggestions

Otherwise, it is fast and accurate if two images are the same orientation.

Thanks

[issa-ai]

This is something expected from a deep learning model. Rotate your images first.

[guipotje]

Please check this issue. You can use this approach to match images of rotated scenes.