Privacy Preserving Ordinal-Meta Learning with VLMs for Fine-Grained Fruit Quality Prediction

2025-11-03

AI Summary

Freshness prediction for fruits and vegetables faces dual challenges of scarce expert annotations and high data privacy sensitivity. Method: We propose a model-agnostic ordinal meta-learning framework that tightly integrates vision-language models (VLMs), ordinal regression, and meta-learning, augmented by a knowledge distillation strategy from proprietary VLMs. Our approach requires no access to raw sensitive data and achieves fine-grained quality assessment using only a minimal number of ordinal-labeled samples (e.g., “fresh → slightly spoiled → severely spoiled”). Contribution/Results: It is the first work to embed ordinal structural priors into the meta-learning paradigm, jointly preserving label semantic order and enabling cross-task generalization. Under zero-shot and few-shot settings, our method achieves a mean accuracy of 92.71%, significantly outperforming existing open-source VLM-based approaches. This provides a scalable, privacy-preserving paradigm for agricultural visual perception under low-data and high-privacy constraints.

Technology Category

• Computer Vision: Large Vision Models
• Machine Learning: Privacy
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Security and Privacy: Data transparency and provenance
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

To effectively manage the wastage of perishable fruits, it is crucial to accurately predict their freshness or shelf life using non-invasive methods that rely on visual data. In this regard, deep learning techniques can offer a viable solution. However, obtaining fine-grained fruit freshness labels from experts is costly, leading to a scarcity of data. Closed proprietary Vision Language Models (VLMs), such as Gemini, have demonstrated strong performance in fruit freshness detection task in both zero-shot and few-shot settings. Nonetheless, food retail organizations are unable to utilize these proprietary models due to concerns related to data privacy, while existing open-source VLMs yield sub-optimal performance for the task. Fine-tuning these open-source models with limited data fails to achieve the performance levels of proprietary models. In this work, we introduce a Model-Agnostic Ordinal Meta-Learning (MAOML) algorithm, designed to train smaller VLMs. This approach utilizes meta-learning to address data sparsity and leverages label ordinality, thereby achieving state-of-the-art performance in the fruit freshness classification task under both zero-shot and few-shot settings. Our method achieves an industry-standard accuracy of 92.71%, averaged across all fruits. Keywords: Fruit Quality Prediction, Vision Language Models, Meta Learning, Ordinal Regression

Problem

• Predicting fruit freshness using visual data with limited expert labels
• Overcoming data privacy concerns with proprietary vision language models
• Improving open-source model performance for fine-grained fruit classification

Innovation

• Meta-learning algorithm trains smaller VLMs
• Utilizes label ordinality for data sparsity
• Achieves state-of-the-art fruit freshness classification