Video Action Recognition on Efficiency Benchmarking (val)
102GFLOPSEZ-CLIP
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| EZ-CLIPHardware=single A100 GPU2023.12 | 102 | 322.4 | 87.5 | 5.2 | |
| ViFi CLIPHardware=single A100 GPU2023.12 | 281 | 71.1 | 124.7 | 124.7 | |
| ViFi-CLIPGPU=A6000 Ada2026.05 | 285 | 43 | 124.3 | — | |
| XCLIPHardware=single A100 GPU2023.12 | 287 | 58.5 | 131.5 | 131.5 | |
| X-CLIPGPU=A6000 Ada2026.05 | 288 | 40 | 169.7 | — | |
| TC-CLIPGPU=A6000 Ada2026.05 | 304 | 27 | 127.5 | — | |
| Vita-CLIPGPU=A6000 Ada2026.05 | 307 | 34 | 161.8 | — | |
| SimVAGPU=A6000 Ada2026.05 | 315 | 29 | 124.6 | — | |
| BDC-CLIPGPU=A6000 Ada2026.05 | 316 | 30 | 126.9 | — | |
| ActionCLIPHardware=single A100 GPU2023.12 | 563 | 67.7 | 168.5 | 168.5 | |
| ActionCLIPGPU=A6000 Ada2026.05 | 567 | 23 | 143.7 | — |