ziplab / LITv2

[NeurIPS 2022 Spotlight] This is the official PyTorch implementation of "Fast Vision Transformers with HiLo Attention"
Apache License 2.0
241 stars 14 forks source link

Training with non-square images #10

Closed ytring closed 1 year ago

ytring commented 1 year ago

Hi, thank you for the great work.

Can you tell how I can train with non-square images? For example, images with size 640 x 480.

HubHop commented 1 year ago

Hi @tonygraf, thanks for your interest! At this moment, most imagenet classfication training scripts are based on squared images. However, it is straightforward to extend the model architecture to handle non-squared images. You can refer to our detection code here for an example. I will try to find some time to improve the training scripts.

ytring commented 1 year ago

Ok. Thank you for your help.