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

Recently AddedMost HypedMost ActiveNeeds VerificationMost Verified

Showing 1351–1375 of 2759 papers

TitleStatusHype
Learning a Deep Model for Human Action Recognition from Novel Viewpoints—0
Learning a discriminative hidden part model for human action recognition—0
Learning and Recognizing Human Action from Skeleton Movement with Deep Residual Neural Networks—0
Learning and Refining of Privileged Information-based RNNs for Action Recognition from Depth Sequences—0
Learning and Using the Arrow of Time—0
Learning a Non-Linear Knowledge Transfer Model for Cross-View Action Recognition—0
Learning a Pose Lexicon for Semantic Action Recognition—0
Learning Chebyshev Basis in Graph Convolutional Networks for Skeleton-based Action Recognition—0
Hyper-Fisher Vectors for Action Recognition—0
HyLiFormer: Hyperbolic Linear Attention for Skeleton-based Human Action Recognition—0
Learning Compact Recurrent Neural Networks with Block-Term Tensor Decomposition—0
Learning Comprehensive Motion Representation for Action Recognition—0
Cycle-Contrast for Self-Supervised Video Representation Learning—0
Learning Connectivity with Graph Convolutional Networks for Skeleton-based Action Recognition—0
Learning Correlation Structures for Vision Transformers—0
Learning Cross-modal Contrastive Features for Video Domain Adaptation—0
AdaSGN: Adapting Joint Number and Model Size for Efficient Skeleton-Based Action Recognition—0
Learning Discriminative Motion Features Through Detection—0
ASL Recognition with Metric-Learning based Lightweight Network—0
Learning Discriminative Spatio-temporal Representations for Semi-supervised Action Recognition—0
HumMUSS: Human Motion Understanding using State Space Models—0
Contrastive Video Representation Learning via Adversarial Perturbations—0
Learning Causal Domain-Invariant Temporal Dynamics for Few-Shot Action Recognition—0
Learning End-to-End Action Interaction by Paired-Embedding Data Augmentation—0
HuMMan: Multi-Modal 4D Human Dataset for Versatile Sensing and Modeling—0
Show:102550
← PrevPage 55 of 111Next →
All datasetsSomething-Something V2UCF101HMDB-51Something-Something V1AVA v2.2EPIC-KITCHENS-100NTU RGB+DNTU RGB+D 120Diving-48ActivityNetAVA v2.1H2O (2 Hands and Objects)

Benchmark Results

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