Reproduce “STPro: Spatial and Temporal Progressive Learning for Weakly Supervised Spatio-Temporal Grounding” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Aaryan Garg, Akash Kumar, Yogesh S Rawat
STPro adapts an image-grounding model to weakly supervised video by linking detections across frames into tubelets and separating temporal action localization from subject selection. It progressively trains on longer ordered action combinations and more varied candidate scenes. On HCSTVG-v1, it reaches 17.56 m_vIoU versus VCMA’s 14.64; on HCSTVG-v2 validation, 19.99 versus 9.85 for W-GDINO.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.