Open-vocabulary object recognition on EPIC100-OV
52.2Top-1 Accuracy (Base)S3D
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| S3DZero-shot evaluation=true2023.08 | 52.2 | 1.6 | 3.1 | |
| AOPBackbone=CLIP ViT-L/142023.08 | 47.8 | 22.6 | 30.7 | |
| AOPBackbone=CLIP ViT-B/162023.08 | 41.9 | 19.2 | 26.3 | |
| C4Backbone=ViT-B/162023.08 | 39.9 | 7.3 | 12.3 | |
| CLIPBackbone=ViT-B/16, Zero-shot evaluation=true2023.08 | 20.2 | 0.6 | 1.2 | |
| C2Backbone=ViT-B/162023.08 | 19.1 | 9.9 | 10.3 | |
| DETIC2023.08 | 12 | 2.3 | 3.9 | |
| CLIPBackbone=ViT-B/16, Zero-shot evaluation=false2023.08 | 7.9 | 16.3 | 10.6 |