Abstract:
"In recent years, tremendous progress has been witnessed in the field of computer vision, particularly concerning object detection and image classification methods. Within this context, vegetable detection systems play a crucial role in various applications, including dietary monitoring and agricultural automation. However, existing methods often face challenges related to accuracy, speed, and adaptability. A novel integration of the Faster RCNN object detection model with a VGG16's image classification has
been presented to address these limitations. The resulting system demonstrates impressive accuracy in identifying vegetables across diverse scenarios, including situations involving occlusion and varying lighting conditions. By leveraging both VGG16 and Faster RCNN, the method achieves faster inference times and improved accuracy. The primary contributions of this work lie in bridging the gap between robust image classification and start-of-the-art object detection. Additionally, beyond vegetable identification, the system
provides recipe suggestions based on the recognized produce. This unique fusion is poised to pave the way for a more versatile and efficient vegetable detection system. The initial version of the vegetable detection system relies on the hybrid approach between the Faster RCNN and VGG16 model architectures. Despite extensive training efforts and computational complexity, this approach achieved 82% accuracy in vegetable identification. Recognizing the need for a more dependable and effective solution in a breakthrough that can enhance the classification between each category."
