How do you process non-square images passed to CLIP when doing inference?

HumanAIGC / AnimateAnyone

Animate Anyone: Consistent and Controllable Image-to-Video Synthesis for Character Animation

Apache License 2.0

14.23k stars 952 forks source link

How do you process non-square images passed to CLIP when doing inference? #30

Open RedRAINXXXX opened 8 months ago

RedRAINXXXX commented 8 months ago

You use square patches during training, but how do you handle non-square images during inference? Directly resize non-square images to (224, 224)? Looking forward to your reply;)