Integrative zero-shot learning for fruit recognition

Dat Tran-Anh

doi:10.1007/s11042-024-18439-x

Integrative zero-shot learning for fruit recognition

2024-02-10

0

PRE

AI

DOI：10.1007/s11042-024-18439-x

Original

Original request for help

Save

Abstract

En 中文

In order to address the pressing issue of detecting and identifying counterfeit agricultural products, which cause significant economic losses, this paper presents a mobile-based system designed to identify rare agricultural products for consumers. To tackle the problem of identifying fake agricultural products, we propose a novel self-monitoring deep learning model named iCZSL. This model combines the Zero-shot segmentation model and Graph embeddings with the calculation of edit distance using a candidate list obtained from a dictionary. To evaluate the effectiveness of our approach, we conduct experiments on a dataset consisting of different types of potatoes, including chinese potatoes, dalat potatoes (a rare agricultural product), and other varieties, referred to as the TLU-states dataset. The experimental results demonstrate the superiority of our proposed method over competing baselines in terms of both quantitative and qualitative performance measures. This system holds promising potential for effectively detecting and differentiating rare agricultural products, thereby mitigating economic losses caused by counterfeit items.

Keywords:

Fruit recognition

Zero-shot learning

Graph convolutional networks

Integrative zero-shot learning for fruit recognition

Abstract

Journal

Organization