DINOv2 (ViT-g/14 frozen model, w/ ViT-Adapter + Mask2former)

Reported on 4 benchmarks across 2 tasks · 1 paper

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Medical2 results

Semantic SegmentationonADE20K
Params (M)· 2023-04-14
1080
best: 3000 (FD-SwinV2-G)
DINOv2: Learning Robust Visual Features without Supervision arXiv:2304.07193
Semantic SegmentationonADE20K
Validation mIoU· 2023-04-14
60.2
best: 63.6 (ViT-P (InternImage-H))
DINOv2: Learning Robust Visual Features without Supervision arXiv:2304.07193

Audio2 results

10-shot image generationonADE20K
Params (M)· 2023-04-14
1080
best: 3000 (FD-SwinV2-G)
DINOv2: Learning Robust Visual Features without Supervision arXiv:2304.07193
10-shot image generationonADE20K
Validation mIoU· 2023-04-14
60.2
best: 63.6 (ViT-P (InternImage-H))
DINOv2: Learning Robust Visual Features without Supervision arXiv:2304.07193