Codesota · Benchmark · Kinetics-400Home/Leaderboards/Kinetics-400
Unknown

Kinetics-400.

Human action recognition across 400 action classes

Paper ↗Leaderboard ↓
§ 01 · SOTA history

Year over year.

§ 02 · Leaderboard

Results by metric.

Found a wrong score or missing run?
Use row edits to send a sourced correction into moderation.
Add / edit result ↗Report issue ↗

Top 1

Top 1 is the reported evaluation metric for Kinetics-400. Codesota tracks published model scores on this metric so readers can compare state-of-the-art results across sources and model families.

Higher is better

Trust tiers for Top 1verifiedpapervendorcommunityunverified

Muted rows were not state of the art when published — an earlier or same-year result already scored better.

RankModelTrustScoreYearLinksFix
01InternVideo2
InternVideo2 (6B parameters). Shanghai AI Lab, 2024. Top-1 on Kinetics-400. Table 2 in paper.
verified92.12024Source ↗Looks wrong?
02VideoMAE V2 (ViT-g)
VideoMAE V2 ViT-g, fine-tuned on Kinetics-400. Top-1 accuracy. Tencent AI Lab, CVPR 2023. Table 1.
verified902023Source ↗Looks wrong?
03ViViT-H
ViViT-H (Video Vision Transformer, Huge). Google, ICCV 2021. Top-1 on Kinetics-400. Table 2.
verified84.92021Source ↗Looks wrong?
04TimeSformer-L
TimeSformer-L (divided space-time attention). Facebook AI, ICML 2021. Top-1 on Kinetics-400. Table 1.
verified80.72021Source ↗Looks wrong?

Accuracy

Accuracy is the reported evaluation metric for Kinetics-400. Codesota tracks published model scores on this metric so readers can compare state-of-the-art results across sources and model families.

Higher is better

Trust tiers for Accuracyverifiedpapervendorcommunityunverified

Muted rows were not state of the art when published — an earlier or same-year result already scored better.

§ 04 · Submit a result

Add to the leaderboard.

← Back to Leaderboards