UTAR Institutional Repository

Modelling Non-Overt Features for Food Origin Recognition

Lim, Jia Hong (2020) Modelling Non-Overt Features for Food Origin Recognition. Final Year Project, UTAR.

[img]
Preview
PDF
Download (14Mb) | Preview

    Abstract

    Good food is differentiated by taste, not by appearance. The rationale of increasing popularity of a food or food provider is inherited from the consumer trusts on their corresponding branding that guarantee great and familiar taste. Despite food are generally distinguishable by their presentation and plating; it is less obvious in the case of simple comfort food as their unique taste are often masked by their highly similar appearances. Two common motivations in identifying food origins are: (1) on the curiosity of where the food comes from and (2) on the trustworthy to confirm the origin of familiar food when generic packaging is used or due to absence of physical labels. This paper designs a food origin classification system using machine-learning to accurately classify local cuisines that is less discriminative with plating cues. The classifier model is constructed using CNN technique and extensive labeling to address the lack of limited discriminative features due to usage of simple and limited ingredients. These approaches are used for training the dataset in order to obtain high accuracy in tracing the food origin. Generally, the system process is divided into two phases, which are data collection and data processing. The experimental result shows that the model is highly accurate with correct detection up to 79% of true positive rate.

    Item Type: Final Year Project / Dissertation / Thesis (Final Year Project)
    Subjects: T Technology > T Technology (General)
    Divisions: Faculty of Information and Communication Technology > Bachelor of Information Systems (Hons) Business Information Systems
    Depositing User: ML Main Library
    Date Deposited: 30 Dec 2020 19:35
    Last Modified: 30 Dec 2020 19:36
    URI: http://eprints.utar.edu.my/id/eprint/3765

    Actions (login required)

    View Item