search

ChineseYjh / SoftCollage

The official PyTorch implementation of the CVPR2022 paper, SoftCollage: A Differentiable Probabilistic Tree Generator for Image Collage
3 stars 2 forks source link

Computing the saliency mask for an image #3

Open foranuj opened 1 year ago

foranuj commented 1 year ago

To retrain these models with a custom set of images, it seems two images are required, one the original actual image and the other that is the masked image that identifies the salient features.

How exactly is that computed and can you provide a utility method for it in the code?

ChineseYjh commented 1 year ago

Thanks for your interest @foranuj Training on custom data doesn't need any masked image.

foranuj commented 1 year ago

Do you have a notebook version of the setup? I am trying to use Google Collab to train,

ChineseYjh commented 1 year ago

Do you have a notebook version of the setup? I am trying to use Google Collab to train,

Nope, python may be more stable for training than notebook.