Zero-shot Video Recognition on UCF, HMDB, and Kinetics-600
74.8HMDB zs AccMoTE
Evaluation results
| Method | Links | ||
|---|---|---|---|
| MoTEEncoder=ViT-L/14, Input Size=16×224, GFLOPs × Views=1299×4×3, Param (M)=346.6, Unified model=true2024.10 | 74.8 | 80.5 | |
| MoTEEncoder=ViT-L/14, Input Size=8×224, GFLOPs × Views=649×4×3, Param (M)=346.6, Unified model=true2024.10 | 74.4 | 80.1 | |
| Open-VCLIPEncoder=ViT-L/14, Input Size=8×224, GFLOPs × Views=/x3x1, Param (M)=304.0, Unified model=true2024.10 | 73.7 | 78.5 | |
| FROSTEREncoder=ViT-B/16, Input Size=8×224, GFLOPs × Views=/x1x1, Param (M)=86.2, Unified model=true2024.10 | 69.1 | 73.7 | |
| DISTEncoder=ViT-L/14, Input Size=8×224, GFLOPs × Views=710×3×1, Param (M)=343.0, Unified model=true2024.10 | 68.1 | 76.4 | |
| MoTEEncoder=ViT-B/16, Input Size=8×224, GFLOPs × Views=141×1×1, Param (M)=98.8, Unified model=true2024.10 | 67.9 | 74.2 | |
| MoTEEncoder=ViT-B/16, Input Size=8×224, GFLOPs × Views=141×4×3, Param (M)=98.8, Unified model=true2024.10 | 67.9 | 74.7 | |
| Open-VCLIPEncoder=ViT-B/16, Input Size=8×224, GFLOPs × Views=/x1x1, Param (M)=86.2, Unified model=true2024.10 | 67.8 | 72.9 | |
| Text4VisEncoder=ViT-L/14, Input Size=8×224, GFLOPs × Views=649×4×3, Param (M)=346.6, Unified model=true2024.10 | 66.6 | 75.3 | |
| ViFi-CLIPEncoder=ViT-B/16, Input Size=16×224, GFLOPs × Views=281×4×3, Param (M)=149.6, Unified model=false2024.10 | 64.4 | 72.9 | |
| Vita-CLIPEncoder=ViT-B/16, Input Size=8×224, GFLOPs × Views=/x1x1, Param (M)=125.0, Unified model=true2024.10 | 61.5 | 69.7 | |
| Vita-CLIPEncoder=ViT-B/16, Input Size=8×224, GFLOPs × Views=/x4x3, Param (M)=125.0, Unified model=true2024.10 | 61.5 | 70.2 | |
| X-FlorenceEncoder=Florence, Input Size=32×224, GFLOPs × Views=2822×4×3, Unified model=false2024.10 | 61.4 | 71.8 | |
| ST-AdapterEncoder=ViT-B/16, Input Size=8×224, GFLOPs × Views=152×3×1, Param (M)=93.0, Unified model=true2024.10 | 61.2 | 70.1 | |
| Text4VisEncoder=ViT-B/16, Input Size=8×224, GFLOPs × Views=141×4×3, Param (M)=105.2, Unified model=true2024.10 | 59.1 | 69 | |
| X-CLIPEncoder=ViT-B/16, Input Size=8×224, GFLOPs × Views=145×1×1, Param (M)=131.5, Unified model=false2024.10 | 58.1 | 68.1 | |
| X-CLIPEncoder=ViT-B/16, Input Size=8×224, GFLOPs × Views=145×4×3, Param (M)=131.5, Unified model=false2024.10 | 58.1 | 68.6 | |
| ActionCLIPEncoder=ViT-B/16, Input Size=16×224, GFLOPs × Views=282×10×3, Param (M)=141.7, Unified model=false2024.10 | 53.2 | 64.7 |