| dc.description.abstract | The advancement of technology in the era of Industry 4.0 digitalization has driven various sectors to increasingly adopt artificial intelligence, including the food industry such as bakery products. This research seeks to design a system that can recognize identifying types and counting the number of bread items rapidly using the Faster Region-Based Convolutional Neural Network (Faster R-CNN) method. The system is designed to address the limitations of time and accuracy in manual counting processes, particularly when dealing with large quantities of bread. Object identification and detection of bread types present unique challenges due to the considerable variation in shape, color, and texture among different types of bread. To overcome these challenges, Faster R-CNN is employed for its ability to simultaneously predict object locations (bounding boxes) and classification labels.
This research utilizes an image dataset of four types of bread (sweet bread, coconut bread, chocolate bread, and floss bread), annotated in the Pascal VOC format. The Faster R-CNN model is trained using the PyTorch framework and integrated into an Android application with a FastAPI backend. The experimental results showed that the model performed reasonably well in detecting and classifying objects, achieving an accuracy of 76.39%, despite several data-related limitations. Nevertheless, the system demonstrated excellent performance in automatically counting objects, even under extreme overlap conditions. | en_US |
