Adaptive Latent Trajectory Anchoring for Action Segmentation Dataset Condensation
2026-07-10 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors improved the way large video datasets for action recognition are compressed by moving from a slow, complex method to a faster, simpler one. They used a new model called Denoising Diffusion Implicit Models to represent video segments efficiently as smooth paths in a special space. They also created a smart system that adjusts how much detail to keep based on how hard each segment is to recreate. Their method works better than previous ones and almost matches training with the original data while using only a small fraction of it.
Dataset CondensationAction SegmentationVariational AutoencoderDenoising Diffusion Implicit ModelsLatent SpaceVideo RepresentationAdaptive AllocationTemporal SegmentationReconstructionCondensation Ratio
Authors
Artheme Gauthier-Villar, Guodong Ding, Angela Yao
Abstract
Dataset condensation for action segmentation synthesizes compact, informative representations of long, untrimmed video datasets. The existing approach relies on Variational Autoencoders and an iterative latent optimization; it is computationally expensive and suffers from over-smoothed reconstructions and rigid temporal constraints. This paper proposes to shift the condensation paradigm from optimization-based inversion to deterministic latent mapping. By leveraging Denoising Diffusion Implicit Models, we represent action segments as continuous trajectories anchored by sparse latent points in the noise manifold. To maximize representational efficiency, we introduce an adaptive allocation mechanism that dynamically redistributes the anchoring budget based on segment-wise reconstruction difficulty. Extensive experiments demonstrate that our framework significantly outperforms state-of-the-art methods in segmentation performance across common datasets. Notably, our approach achieves performance parity with real data training while maintaining a condensation ratio of 2.4\% on Breakfast dataset.