Action recognition: research map
1,325 papers on Action recognition in Video, from ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR (2016–2026), grouped into 2 clusters and 8 approaches. The busiest year so far is 2021.
What it is
Recognising what people are doing in video, from classifying a short clip to finding when each action starts and ends in a long video (temporal action localisation). Inputs include RGB frames and skeleton keypoints.
Terms you will see on this page:
- Temporal action localization
- detecting the start and end time of each action.
- Skeleton-based recognition
- recognising actions from body keypoint sequences.
- Action anticipation
- predicting the next action before it happens.
Start reading with the most cited papers here:
- Quo Vadis, Action Recognition? A New Model and the Kinetics Dataset (CVPR 2017 · 9,707 citations)
- A Closer Look at Spatiotemporal Convolutions for Action Recognition (CVPR 2018 · 3,632 citations)
Within Video, its share shrank from 13.7% in 2023–24 to 5.2% in 2025–26 (141 → 132 papers at ICML, NeurIPS, ICLR, CVPR and AAAI, the venues with data for all four years).
Explore Action recognition in the interactive map
Working on something in this topic? Describe your idea in scime atlas to see which approach it falls under, the closest papers by meaning and how crowded the spot has become.
What changed recently
Approaches whose share of the topic moved most between 2023–24 and 2025–26 (ICML, NeurIPS, ICLR, CVPR and AAAI):
- Disentangled action recognition rising: 23% → 32% of the topic (33 → 42 papers)
- Fully supervised temporal action detection rising: 11% → 13% of the topic (15 → 17 papers)
- Action anticipation and latent actions rising: 21% → 22% of the topic (29 → 29 papers)
- Contrastive video representation learning falling: 11% → 5% of the topic (15 → 6 papers)
- Weakly supervised temporal action localization falling: 11% → 8% of the topic (16 → 11 papers)
Papers from 2025–2026 that are already cited most:
- Revealing Key Details to See Differences: A Novel Prototypical Perspective for Skeleton-based Action Recognition (CVPR 2025 · 36 citations)
Approaches and key papers
Action recognition · 741 papers
Keywords: action recognition, video recognition, representation
Approaches in this cluster:
- Disentangled action recognition (356 papers)
Recognizes actions by decomposing dynamics, components and human-object interactions. - Spatiotemporal CNNs for action recognition (143 papers)
Designs 3D and residual convolutional architectures for video action recognition. - Contrastive video representation learning (132 papers)
Learns video features with contrastive objectives, continuity, motion decoupling and language concepts. - Skeleton-based action recognition (110 papers)
Recognizes actions from skeleton sequences with graph convolutions, LSTMs and prototypes.
Most cited and most cited since 2024:
- Quo Vadis, Action Recognition? A New Model and the Kinetics Dataset (CVPR 2017 · 9,707 citations)
- A Closer Look at Spatiotemporal Convolutions for Action Recognition (CVPR 2018 · 3,632 citations)
- Revealing Key Details to See Differences: A Novel Prototypical Perspective for Skeleton-based Action Recognition (CVPR 2025 · 36 citations)
- ExACT: Language-guided Conceptual Reasoning and Uncertainty Estimation for Event-based Action Recognition and More (CVPR 2024 · 24 citations)
Temporal action localization · 584 papers
Keywords: temporal action, action localization, action detection
Approaches in this cluster:
- Temporal action segmentation (147 papers)
Segments activities in untrimmed videos with unsupervised, weakly or semi-supervised clustering and transformers. - Weakly supervised temporal action localization (147 papers)
Localizes actions using video-level labels with snippet modeling, uncertainty and graph networks. - Fully supervised temporal action detection (159 papers)
Detects action boundaries using temporal convolution, multi-task recycling and context aggregation. - Action anticipation and latent actions (131 papers)
Predicts future actions with LLMs, graph growth and latent action world models.
Most cited and most cited since 2024:
- Temporal Convolutional Networks for Action Segmentation and Detection (CVPR 2017 · 2,175 citations)
- AVA: A Video Dataset of Spatio-Temporally Localized Atomic Visual Actions (CVPR 2018 · 1,021 citations)
- End-to-End Temporal Action Detection with 1B Parameters Across 1000 Frames (CVPR 2024 · 53 citations)
- FACT: Frame-Action Cross-Attention Temporal Modeling for Efficient Action Segmentation (CVPR 2024 · 52 citations)
Related topics in Video
- Video segmentation and tracking (2,850)
- Video-language understanding (1,681)
- Video generation (2,177)
