English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Supervised learning of skeleton sequence encoders for action recognition hasreceived significant attention in recent times. However, learning such encoderswithout labels continues to be a challenging problem. While prior works haveshown promising results by applying contrastive learning to pose sequences, thequality of the learned representations is often observed to be closely tied todata augmentations that are used to craft the positives. However, augmentingpose sequences is a difficult task as the geometric constraints among theskeleton joints need to be enforced to make the augmentations realistic forthat action. In this work, we propose a new contrastive learning approach totrain models for skeleton-based action recognition without labels. Our keycontribution is a simple module, HaLP - to Hallucinate Latent Positives forcontrastive learning. Specifically, HaLP explores the latent space of poses insuitable directions to generate new positives. To this end, we present a noveloptimization formulation to solve for the synthetic positives with an explicitcontrol on their hardness. We propose approximations to the objective, makingthem solvable in closed form with minimal overhead. We show via experimentsthat using these generated positives within a standard contrastive learningframework leads to consistent improvements across benchmarks such as NTU-60,NTU-120, and PKU-II on tasks like linear evaluation, transfer learning, and kNNevaluation. Our code will be made available athttps://github.com/anshulbshah/HaLP.

HaLP: Hallucinating Latent Positives for Skeleton-based Self-Supervised Learning of Actions