TasksSotADatasetsPapersMethodsSubmitAbout
Papers With Code 2

A community resource for machine learning research: papers, code, benchmarks, and state-of-the-art results.

Explore

Notable BenchmarksAll SotADatasetsPapersMethods

Community

Submit ResultsAbout

Data sourced from the PWC Archive (CC-BY-SA 4.0). Built by the community, for the community.

Papers/Make Skeleton-based Action Recognition Model Smaller, Fast...

Make Skeleton-based Action Recognition Model Smaller, Faster and Better

Fan Yang, Sakriani Sakti, Yang Wu, Satoshi Nakamura

2019-07-23arXiv 2019 7Skeleton Based Action RecognitionHand Gesture RecognitionAction Recognition
PaperPDFCodeCode(official)Code

Abstract

Although skeleton-based action recognition has achieved great success in recent years, most of the existing methods may suffer from a large model size and slow execution speed. To alleviate this issue, we analyze skeleton sequence properties to propose a Double-feature Double-motion Network (DD-Net) for skeleton-based action recognition. By using a lightweight network structure (i.e., 0.15 million parameters), DD-Net can reach a super fast speed, as 3,500 FPS on one GPU, or, 2,000 FPS on one CPU. By employing robust features, DD-Net achieves the state-of-the-art performance on our experimental datasets: SHREC (i.e., hand actions) and JHMDB (i.e., body actions). Our code will be released with this paper later.

Results

TaskDatasetMetricValueModel
VideoJHMDB (2D poses only)Average accuracy of 3 splits77.2DD-Net
VideoJ-HMDBAccuracy (pose)77.2DD-Net
Temporal Action LocalizationJHMDB (2D poses only)Average accuracy of 3 splits77.2DD-Net
Temporal Action LocalizationJ-HMDBAccuracy (pose)77.2DD-Net
Zero-Shot LearningJHMDB (2D poses only)Average accuracy of 3 splits77.2DD-Net
Zero-Shot LearningJ-HMDBAccuracy (pose)77.2DD-Net
Activity RecognitionJHMDB (2D poses only)Average accuracy of 3 splits77.2DD-Net
Activity RecognitionJ-HMDBAccuracy (pose)77.2DD-Net
Action LocalizationJHMDB (2D poses only)Average accuracy of 3 splits77.2DD-Net
Action LocalizationJ-HMDBAccuracy (pose)77.2DD-Net
HandDHG-28Accuracy91.9DD-Net
HandSHREC 2017 track on 3D Hand Gesture Recognition14 gestures accuracy94.6DD-Net
HandDHG-14Accuracy94.6DD-Net
Action DetectionJHMDB (2D poses only)Average accuracy of 3 splits77.2DD-Net
Action DetectionJ-HMDBAccuracy (pose)77.2DD-Net
Gesture RecognitionDHG-28Accuracy91.9DD-Net
Gesture RecognitionSHREC 2017 track on 3D Hand Gesture Recognition14 gestures accuracy94.6DD-Net
Gesture RecognitionDHG-14Accuracy94.6DD-Net
3D Action RecognitionJHMDB (2D poses only)Average accuracy of 3 splits77.2DD-Net
3D Action RecognitionJ-HMDBAccuracy (pose)77.2DD-Net
Action RecognitionJHMDB (2D poses only)Average accuracy of 3 splits77.2DD-Net
Action RecognitionJ-HMDBAccuracy (pose)77.2DD-Net

Related Papers

A Real-Time System for Egocentric Hand-Object Interaction Detection in Industrial Domains2025-07-17Visual Hand Gesture Recognition with Deep Learning: A Comprehensive Review of Methods, Datasets, Challenges and Future Research Directions2025-07-06Zero-shot Skeleton-based Action Recognition with Prototype-guided Feature Alignment2025-07-01EgoAdapt: Adaptive Multisensory Distillation and Policy Learning for Efficient Egocentric Perception2025-06-26Feature Hallucination for Self-supervised Action Recognition2025-06-25CARMA: Context-Aware Situational Grounding of Human-Robot Group Interactions by Combining Vision-Language Models with Object and Action Recognition2025-06-25Including Semantic Information via Word Embeddings for Skeleton-based Action Recognition2025-06-23Adapting Vision-Language Models for Evaluating World Models2025-06-22