Why the resize does not preserve the original aspect ratio

[yurymalkov]

Hi, thank you for the work!

I've played with the code and noticed that examples do not preserve the original aspect ration during the resize. E.g. https://github.com/salesforce/BLIP/blob/d10be550b2974e17ea72e74edc7948c9e5eab884/predict.py#L93 or the colab example.

I wonder if it is done on purpose?

[saffie91]

I came to ask the same question, is there no need to do letterboxing?

[LiJunnan1992]

Thanks for your question. It is not necessary to preserve the original aspect ratio because the model was trained on augmented image crops with various ratios. We find this transform to perform slightly better than CenterCrop during inference.