SamsungLabs / ritm_interactive_segmentation

Reviving Iterative Training with Mask Guidance for Interactive Segmentation
MIT License
634 stars 125 forks source link

How to train on my own dataset? #38

Closed Zarxrax closed 10 months ago

Zarxrax commented 10 months ago

I see in config.yaml I can specify paths to some preexisting datasets. But there does not appear to be any commandline argument to specify which one to train on. Does it train on every dataset listed in the config? If I want to train on my own images, how do I do that? And what format does my dataset need to be in? Right now my dataset consists of images + binary image masks. I assume I will have to convert it.