dvlab-research/PFENet

Training the model on binary image segmentation dataset

Ahmedsghir opened this issue · 2 comments

Training the model on binary image segmentation dataset

what is the split 10,11 and 999 ?

@Ahmedsghir
Thank you for being interested in our work.

For the conventional binary segmentation tasks, models for salient object detection and matting might be more effective than our model that is designed for few-shot segmentation. Our model does not update the backbone parameters, which may cause performance deduction on the conventional binary tasks.

Split 10,11 and 999 are deprecated. Sorry for the inconvenience.