Pagri, Pagri (2026) Deteksi Tingkat Kematangan pada Buah Tomat Pasca Panen Menggunakan yolov8 ( studi kasus : Kelompok Tani Desa Puncak Pato Provinsi Sumatera Barat). Other thesis, Universitas Islam Riau.

Text skripsi_213510597_watermark.pdf - Published Version Restricted to Registered users only Download (3MB) | Request a copy

Abstract

This research addresses the primary problem faced by farmers in Puncak Pato Village, where the post-harvest sorting of tomatoes is still performed manually, resulting in inconsistent quality, human errors, and reduced market value. To solve this issue, the study aims to develop an automated system capable of detecting tomato ripeness levels and defects using the YOLOv8 object detection algorithm to achieve more accurate, objective, and efficient sorting. The method includes collecting 550 tomato images under various real-field conditions, annotating six object classes (unripe, turning, half-ripe, ripe, defective, and crown), applying data augmentation methods such as flipping, rotation, noise, and color jitter, and training the model using three dataset split schemes (80:10:10, 70:20:10, and 70:15:15). The YOLOv8 model was trained for 100 epochs and evaluated using Precision, Recall, mAP50, and mAP50–95 metrics. The best performance was achieved with the 80:10:10 split, obtaining an mAP50–95 score of 0.761. Field testing on 1,000 tomatoes resulted in an 78.27% detection success rate, with the highest performance occurring during midday lighting conditions. The system was implemented into a Flask-based web application integrated with smartphone cameras through DroidCam for real-time detection. The main contribution of this research lies in integrating both ripeness classification and defect detection into a single YOLOv8 model and applying it directly in farmers’ real environments, providing a practical and ready-to-use technological solution that improves sorting consistency and supports the post-harvest quality of tomatoes.

Actions (login required)

View Item