Website-Based Boycott Product Detection System using Convolutional Neural Network: A comparison of YOLOv8 and VGG16

Abstrak

Background: In the digital age, product boycotts have become a common form of social protest, but manual product identification is often slow and inaccurate. An automated system is needed to help the public verify product status in real time. Method: This study developed a website-based detection system that compares two deep learning architectures: a one-stage detector (YOLOv8) and a standard classifier (VGG16). The dataset consists of 4,250 images (3,638 boycott products and 612 non-boycott products) collected from the internet. The data was divided into training (70%), validation (20%), and test (10%) sets. YOLOv8 was trained with 640x640 pixel inputs, while VGG16 used 224x224 pixels with transfer learning. Key Results: Testing on 425 images showed that YOLOv8 achieved superior performance with 91.7% accuracy, 99.7% precision, 90.3% recall, and an F1-score of 94.9%. In contrast, VGG16 only achieved an accuracy of 51.3% with low recall (55.7%), indicating difficulty in handling background variations without object localization. Conclusion: YOLOv8 proved to be more effective and efficient than VGG16 for product detection applications on web-based systems. This research provides a practical tool for the public and suggests the integration of OCR features for future development

Identifier

• OpenAlex: https://openalex.org/W7172413429
• DOI: https://doi.org/10.56705/ijodas.v7i2.317