Retraining the SAM model on the color image database in order to segment multiple classes in the imageā€

šŸ¤—Transformers
1

Hello

I wanted to train the SAM (segment anything model) model on my database, my database has 400x600 color images. In each image, there is a ground truth image that specifies the objects and the class of each object. Like the Meta company database.

I wanted to train this pre-trained model on my database for segmentation purposes. But I didnā€™t find a guide on the huggingface.co site. There is an example of retraining this model on ultrasound images where there is only one object with a certain class in each image.

Please guide me.