OMNIVORE (Swin-B)

Reported on 6 benchmarks across 5 tasks · 1 paper · 4 SOTA

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

Computer Vision5 results

Scene ParsingonSUN-RGBD
Accuracy (%)· uses extra data· 2022-01-20
67.2
SOTA
Omnivore: A Single Model for Many Visual Modalities arXiv:2201.08377
AnimationonSUN-RGBD
Accuracy (%)· uses extra data· 2022-01-20
67.2
SOTA
Omnivore: A Single Model for Many Visual Modalities arXiv:2201.08377
3D Character Animation From A Single PhotoonSUN-RGBD
Accuracy (%)· uses extra data· 2022-01-20
67.2
SOTA
Omnivore: A Single Model for Many Visual Modalities arXiv:2201.08377
VideoonKinetics-400
Acc@1· 2022-01-20
84
best: 93.6 (OmniVec2)
Omnivore: A Single Model for Many Visual Modalities arXiv:2201.08377
VideoonKinetics-400
Acc@5· 2022-01-20
96.2
best: 98.9 (TubeViT-H (ImageNet-1k))
Omnivore: A Single Model for Many Visual Modalities arXiv:2201.08377

Audio1 result

2D Semantic SegmentationonSUN-RGBD
Accuracy (%)· uses extra data· 2022-01-20
67.2
SOTA
Omnivore: A Single Model for Many Visual Modalities arXiv:2201.08377