Pineapples Health Detection Using Deep Learning Models

Abstract

This project aims to develop an efficient system for detecting defects in pineapples using advanced deep learning techniques, specifically leveraging YOLO (You Only Look Once) models YOLOv11, YOLOv12, and YOLOv26. The system classifies pineapples as either healthy or defective based on images, addressing the need for an automated quality control solution. The dataset used for training includes images of pineapples exhibiting various defects such as bruising, discoloration, and rot. The deep learning models are trained to recognize these defects, enabling the system to classify the health condition of the pineapples effectively. The project includes a live detection feature, allowing users to capture images and receive immediate feedback. The system is built using Flask and Python for the back-end, while the front-end is developed with HTML, CSS, and JavaScript, creating an intuitive and user-friendly interface. Performance evaluation metrics such as accuracy, precision, recall, and mAP (mean Average Precision) are employed to assess the effectiveness of the models. The project aims to reduce the need for manual inspection, making the defect detection process more efficient, consistent, and reliable, while offering scalability for potential future applications in other agricultural products.

Keywords: Pineapple defect detection, deep learning, YOLO models, object detection, computer vision, dataset pre-processing, image classification, Flask, precision, accuracy.