facebookresearch / GuidedDistillation

Official implementation of the paper "Guided Distillation for Semi-Supervised Instance Segmentation".
Other
35 stars 4 forks source link

About generating the pseudo labels #10

Open antopost opened 5 months ago

antopost commented 5 months ago

Hi, I read your paper and was wondering why you choose to threshold the output of the teacher model and then argmax it, effectively turning the soft labels into hard labels. Screenshot from 2024-06-27 08-53-07

Was it not possible to train with soft labels (i.e. labels with confidence scores of the teacher model) or did thresholding + argmax provide better performance?