Video Classification on Kinetics-600
87.9Top-1 AccuracyMViTv2-L
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| MViTv2-Lpre-train=IN-21K, Resolution=352x352, Input=40x3, FLOPs=3790, Views=3x4, Param=217.62021.12 | 87.9 | 97.9 | — | — | |
| MViTv2-Lpre-train=IN-21K, Resolution=312x312, Input=40x3, FLOPs=2828, Views=3x4, Param=217.62021.12 | 87.5 | 97.8 | — | — | |
| MViTv2-Lpre-train=IN-21K, Resolution=312x312, Input=32x3, FLOPs=2063, Views=3x4, Param=217.62021.12 | 87.2 | 97.6 | — | — | |
| Swin-Lpre-train=IN-21K, Resolution=384x384, FLOPs=2107, Views=5x10, Param=200.02021.12 | 86.1 | 97.3 | — | — | |
| MViTv2-BInput=32x3, FLOPs=206, Views=1x5, Param=51.42021.12 | 85.5 | 97.2 | — | — | |
| MoViNet-A6 + AutoAugmentGFLOPS=386, Resolution=320, Frames=1x120, FPS=12, Parameters=31.4M2021.03 | 84.8 | — | — | — | |
| MoViNet-A6FLOPs=386, Views=1x1, Param=31.42021.12 | 84.8 | 96.5 | — | — | |
| MoViNet-A5 + AutoAugmentGFLOPS=281, Resolution=320, Frames=1x120, FPS=12, Parameters=15.7M2021.03 | 84.3 | — | — | — | |
| SCT-LViews=4×3, #Para=60M2021.08 | 84.3 | 96.3 | — | — | |
| MViTv1-B-24Input=32x3, FLOPs=236, Views=1x5, Param=52.92021.12 | 84.1 | 96.5 | — | — | |
| Swin-BFLOPs=282, Views=3x4, Param=88.12021.12 | 84 | 96.5 | — | — | |
| MoViNet-A6GFLOPS=386, Resolution=320, Frames=1x120, FPS=12, Parameters=31.4M2021.03 | 83.5 | — | — | — | |
| MoViNet-A4 + AutoAugmentGFLOPS=105, Resolution=290, Frames=1x80, FPS=8, Parameters=4.9M2021.03 | 83 | — | — | — | |
| ViViT-L/16x2GFLOPS=3990, Resolution=320, Frames=12x32, FPS=12, Parameters=88.9M2021.03 | 83 | — | — | — | |
| ViViT-LViews=4×3, #Para=89M2021.08 | 83 | 95.7 | — | — | |
| ViT-L-ViViTFLOPs=3992, Views=3x4, Param=310.82021.12 | 83 | 95.7 | — | — | |
| MoViNet-A5GFLOPS=281, Resolution=320, Frames=1x120, FPS=12, Parameters=15.7M2021.03 | 82.7 | — | — | — | |
| TimeSformer-HRGFLOPS=5110, Resolution=224, Frames=3x8, FPS=1.5, Parameters=120M2021.03 | 82.4 | — | — | — | |
| TimeSformerViews=1×3, #Para=121M2021.08 | 82.4 | 96 | — | — | |
| ResNet3D-50-GGFLOPS=3666, Resolution=224, Frames=1x250, FPS=252021.03 | 82 | — | — | — | |
| X3D-XLGFLOPS=1452, Resolution=356, Frames=10x16, FPS=5, Parameters=11.0M2021.03 | 81.9 | — | — | — | |
| X3D-XLViews=10×3, #Para=11M2021.08 | 81.9 | 95.5 | — | — | |
| X3D-XLFLOPs=48.4, Views=3x10, Param=11.02021.12 | 81.9 | 95.5 | — | — | |
| SlowFast-R101GFLOPS=7020, Resolution=256, Frames=30x16, FPS=5, Parameters=59.9M2021.03 | 81.8 | — | — | — | |
| SlowFastViews=10×3, #Para=60M2021.08 | 81.8 | 95.1 | — | — | |
| SlowFast 16x8+NLFLOPs=234, Views=3x10, Param=59.92021.12 | 81.8 | 95.1 | — | — | |
| SCT-MViews=4×3, #Para=33M2021.08 | 81.7 | 95.5 | — | — | |
| LGD-R101Resolution=224, Frames=15x16, FPS=252021.03 | 81.5 | — | — | — | |
| LGD-3D2021.08 | 81.5 | 95.6 | — | — | |
| MoViNet-A3 + AutoAugmentGFLOPS=56.9, Resolution=256, Frames=1x120, FPS=12, Parameters=5.3M2021.03 | 81.3 | — | — | — | |
| MoViNet-A4GFLOPS=105, Resolution=290, Frames=1x80, FPS=8, Parameters=4.9M2021.03 | 81.2 | — | — | — | |
| ResNet3D-152Reproduced=true, GFLOPS=1400, Resolution=224, Frames=1x250, FPS=25, Parameters=80.1M2021.03 | 81.1 | — | — | — | |
| MoViNet-A3GFLOPS=56.9, Resolution=256, Frames=1x120, FPS=12, Parameters=5.3M2021.03 | 80.8 | — | — | — | |
| X3D-LReproduced=true, GFLOPS=187, Resolution=356, Frames=1x120, FPS=2, Parameters=6.1M2021.03 | 80.7 | — | — | — | |
| X3D-LGFLOPS=744, Resolution=356, Frames=30x16, FPS=5, Parameters=6.1M2021.03 | 80.5 | — | — | — | |
| X3D-XLReproduced=true, GFLOPS=151, Resolution=356, Frames=1x50, FPS=5, Parameters=11.0M2021.03 | 80.3 | — | — | — | |
| AttenNAS2021.08 | 79.8 | 94.4 | — | — | |
| Supervised (K600)Backbone=R3D-50 (31.7M), Pre-train data=N/A, Modality=V, Evaluation Protocol=N/A2020.08 | 79.4 | — | — | — | |
| X3D-LReproduced=true, GFLOPS=77.5, Resolution=356, Frames=1x50, FPS=5, Parameters=6.1M2021.03 | 79.1 | — | — | — | |
| X3D-MGFLOPS=186, Resolution=256, Frames=30x16, FPS=5, Parameters=3.8M2021.03 | 78.8 | — | — | — | |
| SlowFast-R50GFLOPS=1080, Resolution=256, Frames=30x16, FPS=5, Parameters=34.4M2021.03 | 78.8 | — | — | — | |
| ResNet3D-50Reproduced=true, GFLOPS=390, Resolution=224, Frames=1x250, FPS=25, Parameters=34.0M2021.03 | 78.7 | — | — | — | |
| MoViNet-A2GFLOPS=10.3, Resolution=224, Frames=1x50, FPS=5, Parameters=4.8M2021.03 | 77.5 | — | — | — | |
| SCT-SViews=4×3, #Para=19M2021.08 | 77.5 | 93.1 | — | — | |
| PyraTokVAE=true2026.01 | 77.11 | — | — | — | |
| X3D-MReproduced=true, GFLOPS=19.4, Resolution=256, Frames=1x50, FPS=5, Parameters=3.8M2021.03 | 76.9 | — | — | — | |
| X3D-SGFLOPS=76.1, Resolution=182, Frames=30x13, FPS=4, Parameters=3.8M2021.03 | 76.4 | — | — | — | |
| PEcoreGVAE=false2026.01 | 76.1 | — | — | — | |
| MoViNet-A1GFLOPS=6.02, Resolution=172, Frames=1x50, FPS=5, Parameters=4.6M2021.03 | 76 | — | — | — | |
| X3D-SReproduced=true, GFLOPS=9.75, Resolution=182, Frames=1x50, FPS=5, Parameters=3.8M2021.03 | 74.3 | — | — | — | |
| NP-MixInput Modality=video and optical flow2025.05 | 73.67 | — | — | — | |
| Feature MixingInput Modality=video and optical flow2025.05 | 73.67 | — | — | — | |
| NPOSInput Modality=video and optical flow2025.05 | 73.49 | — | — | — | |
| X3D-SReproduced=true, GFLOPS=7.8, Resolution=182, Frames=1x40, FPS=4, Parameters=3.8M2021.03 | 73.4 | — | — | — | |
| MixupInput Modality=video and optical flow2025.05 | 73.4 | — | — | — | |
| BaselineInput Modality=video and optical flow2025.05 | 73.14 | — | — | — | |
| VOSInput Modality=video and optical flow2025.05 | 73.07 | — | — | — | |
| CVRLBackbone=R3D-152 (2x) (328.0M), Pre-train data=K600 (44d), Modality=V, Evaluation Protocol=Linear eval on K6002020.08 | 72.9 | — | — | — | |
| InternVideo2VAE=false2026.01 | 72.8 | — | — | — | |
| X3D-XSGFLOPS=23.3, Resolution=182, Frames=30x4, FPS=2, Parameters=3.8M2021.03 | 72.3 | — | — | — | |
| CVRLBackbone=R3D-101 (59.7M), Pre-train data=K600 (44d), Modality=V, Evaluation Protocol=Linear eval on K6002020.08 | 71.6 | — | — | — | |
| I3DGFLOPS=216, Resolution=224, Frames=1x250, FPS=25, Parameters=12M2021.03 | 71.6 | — | — | — | |
| MoViNet-A0GFLOPS=2.71, Resolution=172, Frames=1x50, FPS=5, Parameters=3.1M2021.03 | 71.5 | — | — | — | |
| MobileNetV3-L+TSMReproduced=true, GFLOPS=11, Resolution=224, Frames=1x50, FPS=5, Parameters=5.4M2021.03 | 71.4 | — | — | — | |
| MMVBackbone=TSM-50×2 (93.9M), Pre-train data=AS + HT (16y), Modality=VAT, Evaluation Protocol=Linear eval on K6002020.08 | 70.5 | — | — | — | |
| CVRLBackbone=R3D-50 (31.7M), Pre-train data=K600 (44d), Modality=V, Evaluation Protocol=Linear eval on K6002020.08 | 70.4 | — | — | — | |
| X3D-XSReproduced=true, GFLOPS=3.88, Resolution=182, Frames=1x20, FPS=2, Parameters=3.8M2021.03 | 70.2 | — | — | — | |
| VideoCoCazero-shot=true, evaluation=single view2022.12 | 70.1 | 88.9 | — | — | |
| VideoCoCa (w.o. videos)zero-shot=true, evaluation=single view, continued pretraining=false2022.12 | 70 | 89 | — | — | |
| InternVL-CNumber of frames (#F)=8, Zero-shot=true2023.12 | 68.9 | — | 78.8 | — | |
| InternVLVAE=false2026.01 | 68.9 | — | — | — | |
| X-Florencezero-shot=true, evaluation=single view2022.12 | 68.8 | 88.4 | — | — | |
| LARPVAE=true2026.01 | 68.52 | — | — | — | |
| MobileNetV3-LReproduced=true, GFLOPS=11, Resolution=224, Frames=1x50, FPS=5, Parameters=5.4M2021.03 | 68.1 | — | — | — | |
| SigLIP2-g-optVAE=false2026.01 | 67 | — | — | — | |
| MobileNetV3-S+TSMReproduced=true, GFLOPS=2.8, Resolution=224, Frames=1x50, FPS=5, Parameters=2.5M2021.03 | 65.5 | — | — | — | |
| InternVL-CNumber of frames (#F)=1, Zero-shot=true2023.12 | 65.5 | — | 75.5 | — | |
| X-CLIPzero-shot=true, evaluation=single view2022.12 | 65.2 | 86.1 | — | — | |
| CoCazero-shot=true, evaluation=single view, pooling=frame-level average2022.12 | 65.1 | 87.1 | — | — | |
| SweetTokVAE=true2026.01 | 65.01 | — | — | — | |
| OmniTokenizerVAE=true2026.01 | 62.75 | — | — | — | |
| ViCLIPNumber of frames (#F)=8, Zero-shot=true2023.12 | 62.2 | — | 73.5 | — | |
| LanguageBindSize=Large, Evaluation Protocol=Zero-shot2023.10 | 61.9 | — | — | — | |
| MobileNetV3-SReproduced=true, GFLOPS=2.8, Resolution=224, Frames=1x50, FPS=5, Parameters=2.5M2021.03 | 61.3 | — | — | — | |
| VideoVae+VAE=true2026.01 | 61.27 | — | — | — | |
| MMV-VABackbone=S3D-G (9.1M), Pre-train data=AS + HT (16y), Modality=VA, Evaluation Protocol=Linear eval on K6002020.08 | 59.8 | — | — | — | |
| OpenCLIPSize=Large, Evaluation Protocol=Zero-shot2023.10 | 59 | — | — | — | |
| ImageNet inflatedBackbone=R3D-50 (31.7M), Pre-train data=ImageNet (N/A), Modality=V, Evaluation Protocol=Linear eval on K6002020.08 | 54.7 | — | — | — | |
| SimCLR inflatedBackbone=R3D-50 (31.7M), Pre-train data=K600 (44d), Modality=V, Evaluation Protocol=Linear eval on K6002020.08 | 51.6 | — | — | — | |
| ER-ZSARzero-shot=true, evaluation=single view2022.12 | 42.1 | 73.1 | — | — | |
| DFN5B-CLIP-H/14Zero-shot=true, #Frames=12024.02 | — | — | 65.5 | 67 | |
| DFN5B-CLIP-H/14+Zero-shot=true, #Frames=12024.02 | — | — | 67 | 68.4 | |
| EVA-01-CLIP-g/14Zero-shot=true, #Frames=12024.02 | — | — | 64.5 | 66.2 | |
| EVA-01-CLIP-g/14+Zero-shot=true, #Frames=12024.02 | — | — | 67.3 | 68.8 | |
| EVA-01-CLIP-g+Number of frames (#F)=1, Zero-shot=true2023.12 | — | — | 67 | — | |
| EVA-02-CLIP-E/14+Zero-shot=true, #Frames=12024.02 | — | — | 70 | 72.1 | |
| EVA-02-CLIP-E+Number of frames (#F)=1, Zero-shot=true2023.12 | — | — | 69.3 | — | |
| EVA-CLIP-18BZero-shot=true, #Frames=12024.02 | — | — | 72.9 | 75 | |
| EVA-CLIP-18BZero-shot=true, #Frames=82024.02 | — | — | 79.2 | 79.7 | |
| EVA-CLIP-18BZero-shot=true, #Frames=162024.02 | — | — | 79.4 | 79.8 |