🤖 AI Summary
This work addresses out-of-distribution action detection (ODAD) in untrimmed videos—specifically, the joint optimization of in-distribution (ID)/out-of-distribution (OOD) action classification and spatiotemporal localization under semantic shift. We propose an uncertainty-guided appearance-motion association network. Methodologically, it employs a dual-branch spatiotemporal graph neural network to model cross-object interactions, integrates appearance and motion features via attention-based fusion, and incorporates uncertainty calibration to enhance OOD discrimination robustness. Our key contribution is the first integration of uncertainty estimation into multimodal feature association modeling, enabling co-optimization of dynamic behavioral representation and confidence-aware inference. On two benchmark datasets, our method achieves improvements of 3.2% in ID classification accuracy, 5.8% in OOD detection F1-score, and 4.1% in localization mAP over current state-of-the-art approaches.
📝 Abstract
Out-of-distribution (OOD) detection targets to detect and reject test samples with semantic shifts, to prevent models trained on in-distribution (ID) dataset from producing unreliable predictions. Existing works only extract the appearance features on image datasets, and cannot handle dynamic multimedia scenarios with much motion information. Therefore, we target a more realistic and challenging OOD detection task: OOD action detection (ODAD). Given an untrimmed video, ODAD first classifies the ID actions and recognizes the OOD actions, and then localizes ID and OOD actions. To this end, in this paper, we propose a novel Uncertainty-Guided Appearance-Motion Association Network (UAAN), which explores both appearance features and motion contexts to reason spatial-temporal inter-object interaction for ODAD. Firstly, we design separate appearance and motion branches to extract corresponding appearance-oriented and motion-aspect object representations. In each branch, we construct a spatial-temporal graph to reason appearance-guided and motion-driven inter-object interaction. Then, we design an appearance-motion attention module to fuse the appearance and motion features for final action detection. Experimental results on two challenging datasets show that UAAN beats state-of-the-art methods by a significant margin, illustrating its effectiveness.