analiza aiqbench
This research introduces a method for generating diverse spatiotemporal video data using diffusion models to enhance computer vision tasks in low-data regimes, such as UAV imagery. By leveraging off-the-shelf video foundation models, the approach effectively expands training distributions along spatial and temporal axes that traditional augmentation techniques miss. The study provides actionable guidelines for transferring annotations to synthetic frames while addressing challenges like disocclusion in generated views.
Generative Spatiotemporal Data Augmentation
arXiv:2512.12508v2 Announce Type: replace-cross Abstract: We explore spatiotemporal data augmentation using video foundation models to diversify both camera viewpoints and scene dynamics. Unlike existing approaches based on simple geometric transforms or appearance perturbations, our method leverages off-the-shelf video diffusion models to generate realistic 3D spatial and temporal variations from a given image dataset. Incorporating these synthesized video clips as supplemental training data yields consistent performance gains in low-data settings, such as UAV-captured imagery where annotations are scarce. Beyond empirical improvements, we provide practical guidelines for (i) choosing an appropriate spatiotemporal generative setup, (ii) transferring annotations to synthetic frames, and (iii) addressing disocclusion - regions newly revealed and unlabeled in generated views. Experiments on COCO subsets and UAV-captured datasets show that, when applied judiciously, spatiotemporal augmentation broadens the data distribution along axes underrepresented by traditional and prior generative methods, offering an effective lever for improving model performance in data-scarce regimes.
Abstract:We explore spatiotemporal data augmentation using video foundation models to diversify both camera viewpoints and scene dynamics. Unlike existing approaches based on simple geometric transforms or appearance perturbations, our method leverages off-the-shelf video diffusion models to generate realistic 3D spatial and temporal variations from a given image dataset. Incorporating these synthesized video clips as supplemental training data yields consistent performance gains in low-data settings, such as UAV-captured imagery where annotations are scarce. Beyond empirical improvements, we provide practical guidelines for (i) choosing an appropriate spatiotemporal generative setup, (ii) transferring annotations to synthetic frames, and (iii) addressing disocclusion - regions newly revealed and unlabeled in generated views. Experiments on COCO subsets and UAV-captured datasets show that, when applied judiciously, spatiotemporal augmentation broadens the data distribution along axes underrepresented by traditional and prior generative methods, offering an effective lever for improving model performance in data-scarce regimes.
Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
arXiv:2512.12508 [cs.CV]
(or arXiv:2512.12508v2 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2512.12508 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 242 cuvinte
Subiecte conexe
Self-Augmented Diffusion Guidance for Physics-Informed Generation
arXiv · 28 aug. 2026
UniWAM: Unified World-Action Model
arXiv · 2 oct. 2026
{\tau}: Learning Touch-Augmented Vision-Language-Action Models from Future Visual Supervision
arXiv · 28 iul. 2026
Graph-centric agentic intelligence
Amazon Science · 1 oct. 2026