Detecting Korean Food Using Image using Hierarchical Model

TL;DR

The paper tackles image-based identification of Korean dishes to aid dietary restrictions by proposing a hierarchical model that structures foods into kinds and items. It introduces a two-stage multi-stage transfer learning pipeline based on Yolov8, comparing it against a flat classifier and achieving higher accuracy (88.50% vs 85.32%). The hierarchical approach addresses class imbalance and visual ambiguity, enabling more interpretable and scalable recognition. The work aims to support image-based dietary assessment and proposes future multi-modal extensions to incorporate nutritional data for practical impact.

Abstract

A solution was made available for Korean Food lovers who have dietary restrictions to identify the Korean food before consuming. Just by uploading a clear photo of the dish, people can get to know what they are eating. Image processing techniques together with machine learning helped to come up with this solution.

Figures (3)

• Figure 1: Caption
• Figure 2: YOLO v8 architecture_2023_brief
• Figure 3: Overall Hierarchical Classify Model