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

2026-08-12

AI Summary

This study addresses the challenges of scarce annotations and high inter-individual variability in daily-level freshness estimation of fish from hyperspectral images by introducing few-shot learning to food quality assessment for the first time. Each fish fillet is modeled as an independent few-shot ordinal regression task. The proposed method integrates a cumulative threshold ordinal prediction head with a CORAL-style regression architecture, enhanced by biologically inspired monotonicity constraints and embedding smoothness regularization to ensure temporally coherent freshness predictions. Under a strict evaluation protocol involving unseen fish fillets, the approach achieves a mean absolute error of 1.58 days and a within-two-days accuracy of 72.3% using only three labeled days per fillet, significantly outperforming scalar regression and label distribution baselines.

Abstract

Non-destructive food quality assessment has increasingly benefited from hyperspectral imaging (HSI), which captures spectral signatures linked to biochemical changes during storage. Estimating day-wise freshness, however, remains challenging owing to strong inter-fillet variability and scarce labelled data per product. All existing deep learning approaches for HSI-based freshness prediction operate under full supervision, requiring densely annotated training sets that are costly to obtain at the individual-product level. We introduce, to the best of our knowledge, the first few-shot learning framework for HSI-based food quality estimation. Each fillet defines a distinct episodic task, and a CORAL-style ordinal prediction head captures the ranked nature of freshness progression through cumulative threshold modelling. Biologically grounded monotonicity and embedding smoothness constraints further guide predictions toward plausible trajectories. On a 16-day salmon HSI dataset under a strict unseen-fillet protocol, our method achieves a mean absolute error of 1.58 days and 2-day accuracy of 72.3% with only three labelled days per fillet, substantially outperforming scalar regression and label-distribution baselines under an identical unseen-fillet protocol.

Problem

• Few-Shot Learning
• Ordinal Learning
• Hyperspectral Imaging
• Freshness Estimation
• Food Quality Assessment

Innovation

• few-shot learning
• ordinal regression
• hyperspectral imaging
• food freshness estimation
• monotonicity constraint