Reproduce “Latent Spectral Prompting for Multimodal Hyperspectral Object Tracking” 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

Wangshu Cai, Jun Zhou, Lei Wang, Mohammad Aminul Islam, Xu Han, Kuldip Paliwal

SProTrack addresses hyperspectral object tracking by learning spectral representations separately from temporal tracking. Its self-supervised Spectral-Spatial AutoEncoder learns from full, unlabeled frames, then prompts a frozen RGB tracker whose temporal model carries motion information; trainable projection modules and LoRA adapters provide spectral readout. On HOT2023’s reported aggregate, it achieves 0.906 DP and 0.697 AUC, but does not lead every domain AUC.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.