Action Recognition on Ego-Exo4D
64.06Prediction AccuracyDense
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| DenseP-Ratio (%)=0, GFLOPs=24.159, Architecture=CLIP ViT-B/162026.06 | 64.06 | 0.55 | 85.71 | 0.36 | 97.35 | 66.36 | 63.62 | |
| DenseP-Ratio (%)=0, GFLOPs=7.512, Backbone=CLIP ViT-B/322026.06 | 62.44 | 14.11 | 70.89 | 9.02 | 98.69 | 64.83 | 63.08 | |
| DenseP-Ratio (%)=0, GFLOPs=7.5122026.06 | 58.83 | 16.48 | 60.45 | 10.04 | 98.13 | 62.06 | 59.8 | |
| Rationale-Informed PruningP-Ratio (%)=20, GFLOPs=18.050, Architecture=CLIP ViT-B/162026.06 | 55.05 | 1.76 | 51.52 | 1.02 | 99.38 | 59.38 | 57.47 | |
| Rationale-Informed PruningP-Ratio (%)=20, GFLOPs=5.962, Backbone=CLIP ViT-B/322026.06 | 50.87 | 21.96 | 66.56 | 11.89 | 100 | 55.45 | 54.12 | |
| Rationale-Informed PruningP-Ratio (%)=20, GFLOPs=6.0392026.06 | 49.88 | 17.2 | 51.27 | 9.68 | 98.61 | 57.96 | 55.5 | |
| Rationale-Informed PruningP-Ratio (%)=25, GFLOPs=17.230, Architecture=CLIP ViT-B/162026.06 | 47.27 | 2.84 | 55.56 | 1.49 | 99.32 | 54.52 | 52.27 | |
| Rationale-Informed PruningP-Ratio (%)=25, GFLOPs=5.8322026.06 | 45.88 | 19.73 | 50.43 | 10.63 | 98.55 | 55.71 | 53.11 | |
| Rationale-Informed PruningP-Ratio (%)=25, GFLOPs=5.760, Backbone=CLIP ViT-B/322026.06 | 45.85 | 21.72 | 65.85 | 11.29 | 100 | 53.7 | 51.97 | |
| Rationale-Informed PruningP-Ratio (%)=30, GFLOPs=5.6102026.06 | 44.32 | 20.93 | 48.93 | 10.93 | 97.82 | 54.36 | 51.08 | |
| ECoFLaPP-Ratio (%)=20, GFLOPs=5.903, Backbone=CLIP ViT-B/322026.06 | 42.22 | 16.86 | 67.68 | 8 | 100 | 50.32 | 47.49 | |
| Rationale-Informed PruningP-Ratio (%)=30, GFLOPs=16.414, Architecture=CLIP ViT-B/162026.06 | 40.26 | 3.12 | 46.3 | 1.49 | 98.88 | 50.22 | 47.25 | |
| Rationale-Informed PruningP-Ratio (%)=30, GFLOPs=5.551, Backbone=CLIP ViT-B/322026.06 | 39.96 | 23.36 | 65.14 | 11.05 | 99.37 | 49.37 | 47.01 | |
| ECoFLaPP-Ratio (%)=20, GFLOPs=5.8342026.06 | 38.76 | 19.23 | 47.91 | 9.5 | 94.19 | 50.89 | 46.54 | |
| ECoFLaPP-Ratio (%)=20, GFLOPs=17.704, Architecture=CLIP ViT-B/162026.06 | 37.86 | 0.27 | 25 | 0.12 | 99.87 | 46.76 | 44.8 | |
| ECoFLaPP-Ratio (%)=30, GFLOPs=5.428, Backbone=CLIP ViT-B/322026.06 | 37.05 | 12.57 | 62.25 | 5.62 | 99.2 | 49.77 | 44.32 | |
| Kwon et al.P-Ratio (%)=20, GFLOPs=6.3802026.06 | 36.42 | 21.71 | 46.86 | 8.96 | 82.78 | 47.96 | 34.17 | |
| ECoFLaPP-Ratio (%)=25, GFLOPs=16.784, Architecture=CLIP ViT-B/162026.06 | 35.61 | 1.72 | 33.33 | 0.78 | 98.41 | 51.6 | 44.32 | |
| ECoFLaPP-Ratio (%)=25, GFLOPs=5.693, Backbone=CLIP ViT-B/322026.06 | 35.43 | 13.52 | 53.93 | 5.73 | 98.59 | 47.75 | 41.82 | |
| Kwon et al.P-Ratio (%)=20, GFLOPs=19.700, Architecture=CLIP ViT-B/162026.06 | 32.09 | 0 | 0 | 0 | 96.97 | 43.69 | 38.23 | |
| Kwon et al.P-Ratio (%)=20, GFLOPs=6.380, Backbone=CLIP ViT-B/322026.06 | 30.02 | 22.01 | 43.12 | 6.93 | 93.17 | 36.29 | 29.33 | |
| ECoFLaPP-Ratio (%)=30, GFLOPs=16.349, Architecture=CLIP ViT-B/162026.06 | 27.07 | 2.45 | 60 | 0.9 | 95.74 | 49.2 | 34.95 | |
| ECoFLaPP-Ratio (%)=25, GFLOPs=5.6632026.06 | 26.29 | 23.88 | 39.63 | 7.65 | 88.43 | 36.24 | 28.32 | |
| ECoFLaPP-Ratio (%)=30, GFLOPs=5.5512026.06 | 25.63 | 28.81 | 35.26 | 9.5 | 86.05 | 40.06 | 28.38 | |
| Kwon et al.P-Ratio (%)=25, GFLOPs=6.0962026.06 | 23.53 | 25.69 | 45.06 | 7.89 | 75.48 | 37.09 | 23.18 | |
| Kwon et al.P-Ratio (%)=25, GFLOPs=6.096, Backbone=CLIP ViT-B/322026.06 | 22.87 | 15.72 | 26.29 | 3.64 | 91.49 | 31.42 | 21.21 | |
| Kwon et al.P-Ratio (%)=30, GFLOPs=5.8132026.06 | 13.31 | 36.89 | 45.07 | 6.81 | 25.57 | 20.74 | 4.72 | |
| Kwon et al.P-Ratio (%)=25, GFLOPs=18.582, Architecture=CLIP ViT-B/162026.06 | 9.98 | 0.4 | 100 | 0.06 | 92.34 | 25.88 | 13.68 | |
| Kwon et al.P-Ratio (%)=30, GFLOPs=5.813, Backbone=CLIP ViT-B/322026.06 | 9.98 | 5 | 6.9 | 0.36 | 65 | 20.53 | 4.66 | |
| GMPP-Ratio (%)=25, GFLOPs=6.7512026.06 | 8.65 | 0.43 | 2.87 | 0.06 | 40.98 | 24.96 | 5.68 | |
| GMPP-Ratio (%)=20, GFLOPs=6.8762026.06 | 8.38 | 2.48 | 5.43 | 0.36 | 18.53 | 11.67 | 2.69 | |
| GMPP-Ratio (%)=30, GFLOPs=6.6292026.06 | 8.23 | 0 | 0 | 0 | 3.08 | 6.33 | 0.42 | |
| OMPP-Ratio (%)=30, GFLOPs=5.2582026.06 | 5.2 | 0 | 0 | 0 | 0 | 0 | 0 | |
| GMPP-Ratio (%)=25, GFLOPs=6.750, Backbone=CLIP ViT-B/322026.06 | 3.82 | 8.47 | 2.35 | 0.6 | 72.88 | 12.97 | 5.14 | |
| OMPP-Ratio (%)=25, GFLOPs=5.6332026.06 | 3.4 | 78.26 | 5.18 | 1.08 | 0 | 0 | 0 | |
| OMPP-Ratio (%)=20, GFLOPs=6.0092026.06 | 3.25 | 86.36 | 5.82 | 1.14 | 0 | 0 | 0 | |
| GMPP-Ratio (%)=20, GFLOPs=6.875, Backbone=CLIP ViT-B/322026.06 | 2.88 | 21.62 | 4.88 | 0.96 | 79.73 | 13.11 | 3.52 | |
| OMPP-Ratio (%)=25, GFLOPs=5.633, Backbone=CLIP ViT-B/322026.06 | 2.13 | 14.58 | 2.11 | 0.42 | 100 | 7.19 | 2.87 | |
| GMPP-Ratio (%)=20, GFLOPs=21.697, Architecture=CLIP ViT-B/162026.06 | 1.98 | 0 | 0 | 0 | 100 | 75 | 0.36 | |
| GMPP-Ratio (%)=30, GFLOPs=6.629, Backbone=CLIP ViT-B/322026.06 | 1.65 | 0 | 0 | 0 | 100 | 5.07 | 2.51 | |
| OMPP-Ratio (%)=20, GFLOPs=6.009, Backbone=CLIP ViT-B/322026.06 | 1.05 | 12 | 0.95 | 0.18 | 96 | 5.25 | 1.43 | |
| Kwon et al.P-Ratio (%)=30, GFLOPs=17.463, Architecture=CLIP ViT-B/162026.06 | 0.66 | 0 | 0 | 0 | 100 | 29.58 | 1.25 | |
| OMPP-Ratio (%)=20, GFLOPs=19.326, Architecture=CLIP ViT-B/162026.06 | 0.48 | 0 | 0 | 0 | 0 | 0 | 0 | |
| OMPP-Ratio (%)=25, GFLOPs=18.117, Architecture=CLIP ViT-B/162026.06 | 0.48 | 0 | 0 | 0 | 0 | 0 | 0 | |
| GMPP-Ratio (%)=25, GFLOPs=21.203, Architecture=CLIP ViT-B/162026.06 | 0.48 | 0 | 0 | 0 | 0 | 0 | 0 | |
| OMPP-Ratio (%)=30, GFLOPs=16.908, Architecture=CLIP ViT-B/162026.06 | 0.48 | 0 | 0 | 0 | 0 | 0 | 0 | |
| GMPP-Ratio (%)=30, GFLOPs=20.706, Architecture=CLIP ViT-B/162026.06 | 0.48 | 0 | 0 | 0 | 0 | 0 | 0 | |
| OMPP-Ratio (%)=30, GFLOPs=5.257, Backbone=CLIP ViT-B/322026.06 | 0.48 | 0 | 0 | 0 | 0 | 0 | 0 |