'Super resolution GAN, use image with different shapes for training
Good morning, I am trying to train a SRGAN model using keras/tensorflow. I was able to run and train the model using high resolution images of the same shape (48x48) and low resolution image how the same shape (12x12). What I am trying to do now is to train the model using high resolution images of different shapes and low resolution images downsampled of a factor of 4. For example, if HR images is 100x100 the corresponding LR image will be 25x25, if another HR image is 60x60 the corresponding LR image will be 15x15 and so on. However, when I run the code I receive this error: ValueError: Failed to convert a NumPy array to a Tensor (Unsupported object type numpy.ndarray). I read online that this can be caused by the fact the HR images have to have the same size. Someone knows if there is a way to train an SRGAN model using images of different size?
Thank you very much
Sources
This article follows the attribution requirements of Stack Overflow and is licensed under CC BY-SA 3.0.
Source: Stack Overflow
| Solution | Source |
|---|