UniverSeg v1 model weights
The model has 64 channel features throghout the architecture and was trained with a support size of 64 image-label pairs.
STA means that the models was trained:
- (S) Synthetic Data
- (T) Task augmentations
- (A) dat Augmentations