An unsupervised GNN-based system reaches 82.66% mAP on the DSV Diving dataset by detecting curvature inflections of learned embedding norms, matching supervised baselines.
During training, we use a rolling window size of W = 7 and Gaussian noise standard deviation of σ = 0.1 to generate noisy input sub-pose sequences
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
UTAL-GNN: Unsupervised Temporal Action Localization using Graph Neural Networks
An unsupervised GNN-based system reaches 82.66% mAP on the DSV Diving dataset by detecting curvature inflections of learned embedding norms, matching supervised baselines.