Frozen Backbone, SwinV2-G-ext22K (HTC)

Reported on 6 benchmarks across 6 tasks · 1 paper

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

Methodology4 results

3DonCOCO minival
box AP· 2022-11-03
59.3
best: 66 (PE_spatial (DETA))
Could Giant Pretrained Image Models Extract Universal Representations?arXiv:2211.02043
2D ClassificationonCOCO minival
box AP· 2022-11-03
59.3
best: 66 (PE_spatial (DETA))
Could Giant Pretrained Image Models Extract Universal Representations?arXiv:2211.02043
2D Object DetectiononCOCO minival
box AP· 2022-11-03
59.3
best: 66 (PE_spatial (DETA))
Could Giant Pretrained Image Models Extract Universal Representations?arXiv:2211.02043
16konCOCO minival
box AP· 2022-11-03
59.3
best: 66 (PE_spatial (DETA))
Could Giant Pretrained Image Models Extract Universal Representations?arXiv:2211.02043

Computer Vision2 results

Object DetectiononCOCO minival
box AP· 2022-11-03
59.3
best: 66 (PE_spatial (DETA))
Could Giant Pretrained Image Models Extract Universal Representations?arXiv:2211.02043
Instance SegmentationonCOCO minival
mask AP· 2022-11-03
51.6
best: 56.6 (Co-DETR)
Could Giant Pretrained Image Models Extract Universal Representations?arXiv:2211.02043