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

Riddhi Jain, Manasi Patwardhan, Aayush Mishra, Parijat Deshpande, Beena Rai · 2025

Read Privacy Preserving Ordinal-Meta Learning with VLMs for Fine-Grained Fruit Quality Prediction by Riddhi Jain, Manasi Patwardhan, Aayush Mishra, Parijat Deshpande, Beena Rai (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-01449.txt.

Description

Riddhi Jain and colleagues judge fruit freshness from photographs, a non-invasive way to cut waste of perishable produce. Fine grained freshness labels from experts are costly, so training data are scarce. Proprietary vision language models do well at the task, but food retailers cannot send their data to closed services, and open models lag behind.

Their algorithm, model agnostic ordinal meta learning, trains smaller vision language models by meta learning across tasks while exploiting the ordered nature of freshness grades. It is designed for both zero shot and few shot use under data sparsity. The reported accuracy, averaged across fruits, is about ninety two point seven percent, which the authors describe as industry standard, with state of the art results for this classification task.

Catalog record

CREATORRiddhi Jain, Manasi Patwardhan, Aayush Mishra, Parijat Deshpande, Beena Rai
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE31 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2511-01449

Browse the full library · About the Tower of Babel