Machine vision-based pepper breed classification Using yolov10 and transfer learning models

Abstract

This study presents a deep learning–based framework for classifying ten varieties of peppers commonly grown in Bangladesh, including capsicum, local chili, Shimla, and Bombay chili. A balanced dataset of 1,000 images was created to ensure fair evaluation across all classes. Three deep learning models—VGG16, ResNet50, and YOLOv10—were evaluated for pepper breed classification. Experimental results show that ResNet50 and VGG16 achieved the highest classification accuracy of 96%, while YOLOv10 achieved 94.25% accuracy with faster real-time inference capability. The findings demonstrate the effectiveness of deep learning for automated pepper classification and highlight its potential applications in agriculture, quality control, and market management.