AI Paper Atlas

Research map / Label-efficient and robust learning / Self-supervised and contrastive

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

Anshul Shah, Aniket Roy, Ketul Shah, Shlok Mishra, David Jacobs, Anoop Cherian, Rama Chellappa

CVPR 2023 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Contrastive versus non-contrastive theory · 38 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Add latent hallucinated positives to single-modality contrastive trainingMoCo skeleton baseline · NTU-60 x-sub linear evaluationaccuracy: +2.3%Improved
Add latent hallucinated positives to CMD multi-modal trainingCMD · NTU-60 x-sub linear evaluationaccuracy: 79.8 -> 82.1Improved
Add latent hallucinated positives to single-modality contrastive trainingMoCo skeleton baseline · NTU-120 x-sub linear evaluationaccuracy: +2.3%Improved
Add latent hallucinated positives to single-modality contrastive trainingMoCo skeleton baseline · PKU-II x-sub linear evaluationaccuracy: +4.5%Improved
Use query-key geodesic interpolation to generate positives (ablation)HaLP · NTU-60 x-sub linear evaluationaccuracy: 77.92%Hurt
Use only the hardest positives at t_c = t* instead of a range of hardness (ablation)HaLP · NTU-60 x-sub linear evaluationaccuracy: 79.4%Hurt
Use a query feature as the hallucination anchor instead of a key feature (ablation)HaLP · NTU-60 x-sub linear evaluationaccuracy: 77.9%Hurt
Use 20 prototypes instead of 10, 40, or 60 (ablation)HaLP · NTU-60 x-sub linear evaluationaccuracy: 79.5 -> 79.7Improved
Replace the rank filter with variant-1 filtering (ablation)HaLP · NTU-60 x-sub linear evaluationaccuracy: 79.7 -> 79.5Hurt
Replace the rank filter with variant-2 filtering (ablation)HaLP · NTU-60 x-sub linear evaluationaccuracy: 79.7 -> 79.6Hurt
Increase the HaLP loss weight from μ=0.5 to μ=1.0 (ablation)HaLP · NTU-60 x-sub linear evaluationaccuracy: 78.8 -> 79.7Improved
Generate 100 positives rather than 0 (ablation)HaLP · NTU-60 x-sub linear evaluationaccuracy: 77.96 -> 79.71Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy