Few-Shot Ordinal Learning for Day-Wise Freshness Estimation with Hyperspectral Fish Images

A new research paper introduces a few-shot learning framework for estimating day-wise freshness of fish using hyperspectral imaging (HSI). The study addresses the challenge of non-destructive food quality assessment, where HSI captures spectral signatures linked to biochemical changes during storage. Traditional deep learning methods require densely annotated training sets, which are costly to obtain at the individual-product level. The proposed framework treats each fish fillet as a distinct episodic task and employs a CORAL-style ordinal prediction head to model the ranked nature of freshness progression through cumulative threshold modeling. Biologically grounded monotonicity and embedding smoothness constraints guide predictions toward plausible trajectories. This approach is the first few-shot learning framework for HSI-based food quality estimation, potentially reducing the need for extensive labeled data. The paper is available on arXiv under the identifier 2608.12230.