Page 52 - THEi Student Applied Research Presentations 2025
P. 52
"AImeOut": ENHANCED SELF-SERVICE
CHECKOUT SYSTEM
Poster
WONG Chun Yan
BSc (Hons) in Information and Communications Technology
Department of Digital Innovation and Technology
OBJECTIVES RESEARCH BACKGROUND
The "AImeOut" Android app aims to Traditional self-service checkout systems rely on barcode scanning, which is
revolutionize self-service checkouts time-consuming, error-prone, and inefficient for non-barcoded items like fruits
by enabling customers to photograph and vegetables. Inspired by camera-based systems like A-1 Bakery’s, which
items for automatic identification and still require staff assistance, this project leverages advances in computer
bill generation, eliminating barcode vision, particularly YOLOv8’s real-time object detection, to develop a fully
dependency, lining up and reliance autonomous, mobile-based solution. The ubiquity of smartphones and open-
on checkout machines. A reasonable source AI libraries supports the development of an Android app to enhance
approach and budget replacement of retail efficiency and customer convenience.
Amazon Go digitalizing. “AImeOut”
seeks to achieve high-accuracy multi- METHODOLOGY
item detection, support store-specific
AI models, integrate e-payment The project utilized YOLOv8 for real-time object detection, trained on the
solutions, and streamline dataset dataset with 76,539 images across 200 classes, restructured to include 30,000
preprocessing for robust performance multi-item images. Customised scripts which converted COCO annotations to
in diverse retail scenarios. YOLO format, resized images to 640x640, and augmented data. The Android
app, built in with Android Studio, supports image capture, item detection, CSV
bill generation, and AlipayHK integration. IoU-based tracking and model-
switching features were also implemented, with video-to-dataset conversion
ABOUT THE INVESTIGATOR using 360-degree turntables for dataset expansion.
As a passionate and innovative IT
fresh graduate. Specializing in Android FINDINGS
application development, I thrive on
transforming creative ideas into unique, AImeOut achieved exceptional detection accuracy (mAP50: 0.995, mAP50-
impactful solutions. My final year 95: 0.899) (The average of the mean average precision calculated at varying
project, AImeOut, exemplifies this by IoU thresholds, ranging from 0.50 and 0.50 to 0.95) post-dataset restructuring,
revolutionizing self-service checkouts enabling robust multi-item recognition. The app’s user-friendly interface, model-
with a YOLOv8-powered Android app that switching capability, and AlipayHK integration streamlined checkouts, reducing
enables barcode-free item detection and staff dependency. Custom scripts enhanced dataset preprocessing efficiency,
seamless e-payment integration. With
hands-on experience from developing while video-based dataset creation proved effective. However, retraining for
RFID and label-printing apps during my new classes yielded suboptimal results due to dataset imbalance. Future efforts
internship. I excel in crafting user-centric will focus on addressing security concerns, optimizing the efficiency of AI model
applications. My knack for generating retraining, and enhancing the system's ability to recognize unrelated items.
novel ideas, combined with strong
technical skills, critical thinking allows
me to deliver exceptional IT solutions.
I am eager to contribute my expertise
and creativity to dynamic teams, driving
innovation in diverse tech projects. My
FYP supervisor is Dr CHEONG Kai Yuen.
41 Student Applied Research Presentations 2025 Student Applied Research Presentations 2025
