Instance Segmentation on COCO (0.1% labeled data)
13.6Mask APGuided Distillation (Ours)
Evaluation results
| Method | Links | |
|---|---|---|
| Guided Distillation (Ours)Backbone=ViT-L, Pre-training=DINOv2, Learning Paradigm=Semi-supervised2023.08 | 13.6 | |
| Mask2FormerBackbone=ViT-L, Pre-training=DINOv2, Learning Paradigm=Supervised2023.08 | 10.2 | |
| Guided Distillation (Ours)Backbone=Swin-L, Pre-training=IN-21k, Learning Paradigm=Semi-supervised2023.08 | 5.8 | |
| Mask2FormerBackbone=Swin-L, Pre-training=IN-21k, Learning Paradigm=Supervised2023.08 | 5.3 |