DeiT-L

Reported on 4 benchmarks across 3 tasks · 2 papers

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

Computer Vision2 results

Image ClassificationonObjectNet
Top-1 Accuracy· uses extra data· 2022-06-02
31.4
best: 82.7 (CoCa)
Optimizing Relevance Maps of Vision Transformers Improves Robustness arXiv:2206.01161
Image ClassificationonObjectNet
Top-5 Accuracy· uses extra data· 2022-06-02
48.5
best: 82.1 (ViT-H/14)
Optimizing Relevance Maps of Vision Transformers Improves Robustness arXiv:2206.01161

Medical1 result

Semantic SegmentationonADE20K val
mIoU· 2022-04-14
55.6
best: 62.8 (BEiT-3)
DeiT III: Revenge of the ViT arXiv:2204.07118

Audio1 result

10-shot image generationonADE20K val
mIoU· 2022-04-14
55.6
best: 62.8 (BEiT-3)
DeiT III: Revenge of the ViT arXiv:2204.07118