Hi! Thanks very much for the repo, I've enjoyed playing around with it.
Noticed that the new HRNet+OCR implementation does not allow for input image size to be set on initialisation. I propose the simple fix below which is in keeping with other models in the repo.
Issue: Model will not compile owing to "HEIGHT" and "WIDTH" not being initialised.
Fix: Add height and width as class attributes which set input layer dimensions in model class method.
Hi! Thanks very much for the repo, I've enjoyed playing around with it.
Noticed that the new HRNet+OCR implementation does not allow for input image size to be set on initialisation. I propose the simple fix below which is in keeping with other models in the repo.
Issue: Model will not compile owing to "HEIGHT" and "WIDTH" not being initialised. Fix: Add height and width as class attributes which set input layer dimensions in model class method.
Thoughts/alternatives welcomed. Cheers!