Skeleton-Based Action Recognition Using Multibranch Adaptive Graph Convolutional Network With Pose Refinement
Bibliographic record
Abstract
A multibranch adaptive graph convolutional network is proposed for human action recognition by combining graph convolutional networks (GCNs), adaptive learning, and multibranch feature extraction. Through the adaptive graph convolution module, this method can adaptively change parameters during the training process, thereby enhancing the flexibility of the model. Furthermore, the integration of shallow-level features (skeleton joints), with deep-level features including skeleton information, motion information, and motion difference information allows our model to capture both spatial and temporal dynamics of human actions, leading to a more comprehensive representation of human action features. The introduction of the spatio-temporal attention mechanism enables our model to focus on key frames and skeleton joints. The attitude correction module makes the input data to the network more reasonable and reduces the interference of noise. The inclusion of the adaptive mechanism makes the network no longer limited to the inherent physical connections, and the flexibility of the network is enhanced. The addition of second-order features makes the features of the skeletal data fully exploited. This attention mechanism enhances the discriminative ability of the model and improves its ability to recognize subtle variations and important cues in human actions. Through experiments on benchmark datasets, NTU-RGB-D and Kinetics-400, our method achieves significant improvements in action recognition performance compared with existing approaches. On the Kinetics-400 dataset, we achieved 36.5% and 59.6% recognition rates under the Top-1 and Top-5 evaluation metrics, respectively, which is an improvement of about 1% compared with the state-of-the-art method. On the NTU-RGB-D dataset, we achieved 95.8% and 89.4% recognition rates under the X-view and X-subject modes, respectively, with excellent results. These results validate the effectiveness of the multi-branch adaptive graph convolutional network for human action recognition tasks.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".