Video Classification on Kinetics-400 Supervised (train)
84.7Top-1 AccX-CLIP-B/16
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| X-CLIP-B/16Pre-training=CLIP-400M, Finetuning=true, Frames=16, Views=4 x 3, GFLOPS=287, Zero-shot=true2023.04 | 84.7 | 96.8 | |
| ActionCLIP-B/16Pre-training=CLIP-400M, Finetuning=true, Frames=32, Views=10 x 3, GFLOPS=563, Zero-shot=false2023.04 | 83.8 | 96.2 | |
| X-CLIP-B/16Pre-training=CLIP-400M, Finetuning=true, Frames=8, Views=4 x 3, GFLOPS=145, Zero-shot=true2023.04 | 83.8 | 96.7 | |
| Uniformer-BPre-training=IN-1k, Finetuning=true, Frames=32, Views=4 x 3, GFLOPS=259, Zero-shot=false2023.04 | 83 | 95.4 | |
| MViTv2-BPre-training=None, Finetuning=true, Frames=32, Views=5 x 1, GFLOPS=225, Zero-shot=false2023.04 | 82.9 | 95.7 | |
| EVL B/16Pre-training=CLIP-400M, Finetuning=false, Frames=8, Views=1 x 3, GFLOPS=444, Zero-shot=false2023.04 | 82.9 | — | |
| Vita-CLIP B/16 (Mc = 8, Ms = 8)Pre-training=CLIP-400M, Finetuning=false, Frames=16, Views=4 x 3, GFLOPS=190, Zero-shot=true2023.04 | 82.9 | 96.3 | |
| Swin-BPre-training=IN-21k, Finetuning=true, Frames=32, Views=4 x 3, GFLOPS=282, Zero-shot=false2023.04 | 82.7 | 95.5 | |
| X-CLIP-B/16Pre-training=CLIP-400M, Finetuning=true, Frames=8, Views=1 x 1, GFLOPS=145, Zero-shot=true2023.04 | 82.3 | 95.4 | |
| Vita-CLIP B/16 (Mc = 8, Ms = 8)Pre-training=CLIP-400M, Finetuning=false, Frames=8, Views=4 x 3, GFLOPS=97, Zero-shot=true2023.04 | 81.8 | 96 | |
| MViTv1-B, 64x3Pre-training=None, Finetuning=true, Frames=64, Views=3 x 3, GFLOPS=455, Zero-shot=false2023.04 | 81.2 | 95.1 | |
| Swin-BPre-training=IN-1k, Finetuning=true, Frames=32, Views=4 x 3, GFLOPS=282, Zero-shot=false2023.04 | 80.6 | 94.6 | |
| Vita-CLIP B/16 (Mc = 8, Ms = 8)Pre-training=CLIP-400M, Finetuning=false, Frames=8, Views=1 x 1, GFLOPS=97, Zero-shot=true2023.04 | 80.5 | 95.9 | |
| MformerPre-training=IN-21k, Finetuning=true, Frames=16, Views=10 x 3, GFLOPS=370, Zero-shot=false2023.04 | 79.7 | 94.2 | |
| TimeSformerPre-training=IN-21k, Finetuning=true, Frames=96, Views=1 x 3, GFLOPS=590, Zero-shot=false2023.04 | 78 | 93.7 | |
| A6Pre-training=CLIP-400M, Finetuning=false, Frames=16, Zero-shot=true2023.04 | 76.9 | 93.5 |