Transfer Learning on Inception ResNet V2 For Expiry Reminder: A Mobile Application Development

• المؤلف الرئيسي: Ong, Wi Yi
• التنسيق: Final Year Project / Dissertation / Thesis
• منشور في: 2020
• الموضوعات: QA76 Computer software
• الوصول للمادة أونلاين: http://eprints.utar.edu.my/4094/1/1801192_FYP_Report_%2D_WI_YI_ONG.pdf; http://eprints.utar.edu.my/4094/

The expiry date is a very common information that every product has. It represents the recommended period of time to use the product or the duration of time to use the product. However, it is hard to keep track of the expiry date when there are a lot of them. The project aims to solve the problem by using deep learning, Optical character Recognition (OCR) and also a smart and iterative user interface to help users record and track expiry dates fast and efficient. A Convolutional Neural Network (CNN), Inception ResNet V2 is trained with a newly created near reality synthetic expiry dates image dataset to recognize and capture the expiry date on products. The Inception ResNet V2 has achieved an accuracy of 0.9964 on synthetic data and an accuracy of 0.9612 on noisy reality data. The success in training and deploying the Inception ResNet v2 into the mobile application can significantly speed up the process of recording down new expiry dates and helps users track the expiry date efficiently. Moreover, various kind of alerts including push notifications, SMS notifications and email notifications are provided to reach and remind user more effectively than a normal application.