SOTAVerified

Action Recognition

Action Recognition is a computer vision task that involves recognizing human actions in videos or images. The goal is to classify and categorize the actions being performed in the video or image into a predefined set of action classes.

In the video domain, it is an open question whether training an action classification network on a sufficiently large dataset, will give a similar boost in performance when applied to a different temporal task or dataset. The challenges of building video datasets has meant that most popular benchmarks for action recognition are small, having on the order of 10k videos.

Please note some benchmarks may be located in the Action Classification or Video Classification tasks, e.g. Kinetics-400.

Papers

Showing 110 of 2759 papers

TitleStatusHype
A Real-Time System for Egocentric Hand-Object Interaction Detection in Industrial Domains0
Zero-shot Skeleton-based Action Recognition with Prototype-guided Feature AlignmentCode1
EgoAdapt: Adaptive Multisensory Distillation and Policy Learning for Efficient Egocentric Perception0
Feature Hallucination for Self-supervised Action Recognition0
CARMA: Context-Aware Situational Grounding of Human-Robot Group Interactions by Combining Vision-Language Models with Object and Action Recognition0
Including Semantic Information via Word Embeddings for Skeleton-based Action Recognition0
Adapting Vision-Language Models for Evaluating World Models0
Active Multimodal Distillation for Few-shot Action Recognition0
Time-Unified Diffusion Policy with Action Discrimination for Robotic Manipulation0
An Effective End-to-End Solution for Multimodal Action Recognition0
Show:102550
← PrevPage 1 of 276Next →

Benchmark Results

#ModelMetricClaimedVerifiedStatus
1MViTv2-B (IN-21K + Kinetics400 pretrain)Top-5 Accuracy93.4Unverified
2RSANet-R50 (8+16 frames, ImageNet pretrained, 2 clips)Top-5 Accuracy91.1Unverified
3MVD (Kinetics400 pretrain, ViT-H, 16 frame)Top-1 Accuracy77.3Unverified
4DejaVidTop-1 Accuracy77.2Unverified
5InternVideoTop-1 Accuracy77.2Unverified
6InternVideo2-1BTop-1 Accuracy77.1Unverified
7VideoMAE V2-gTop-1 Accuracy77Unverified
8MVD (Kinetics400 pretrain, ViT-L, 16 frame)Top-1 Accuracy76.7Unverified
9Hiera-L (no extra data)Top-1 Accuracy76.5Unverified
10TubeViT-LTop-1 Accuracy76.1Unverified
#ModelMetricClaimedVerifiedStatus
1FTP-UniFormerV2-L/143-fold Accuracy99.7Unverified
2OmniVec23-fold Accuracy99.6Unverified
3VideoMAE V2-g3-fold Accuracy99.6Unverified
4OmniVec3-fold Accuracy99.6Unverified
5BIKE3-fold Accuracy98.8Unverified
6SMART3-fold Accuracy98.64Unverified
7OmniSource (SlowOnly-8x8-R101-RGB + I3D-Flow)3-fold Accuracy98.6Unverified
8PERF-Net (multi-distilled S3D)3-fold Accuracy98.6Unverified
9ZeroI2V ViT-L/143-fold Accuracy98.6Unverified
10LGD-3D Two-stream3-fold Accuracy98.2Unverified