Scholarlink

Dragon fruit, mostly called pitaya, is in the cactus family and Cactaceae, indigenous to Mexico’s deserts, Central America, and South America. Since it hit the market, dragon fruit has moved quickly and achieved profits locally and internationally. This study evaluated the performance of five Convolutional Neural Network (CNN) models in classifying the quality and maturity of dragon fruit. The models included MobileNetV3 (small and large), EfficientNetV2-S, NasNet Mobile, and DenseNet-121. Among the evaluated models, NasNet Mobile was the most accurate for maturity grading, achieving an accuracy of 96.64%. On the other hand, EfficientNetV2-S achieved the highest performance for quality classification, with an accuracy of 98.79%. These lightweight CNN models were successfully integrated into a mobile application, PitAIya, designed to classify the quality and maturity of dragon fruit. This research provides a promising direction for future improvements in dragon fruit farming by offering a time-effective and accurate way of assessing quality and maturity.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

PitAIya: A Dragon Fruit Maturity Recognition and Quality Grading Based on Lightweight Convolutional Neural Networks

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Vince Apolinares,
  • Piolo Kyle Catipay,
  • Patrick Paul Lim,
  • Hermoso Tupas

摘要

Dragon fruit, mostly called pitaya, is in the cactus family and Cactaceae, indigenous to Mexico’s deserts, Central America, and South America. Since it hit the market, dragon fruit has moved quickly and achieved profits locally and internationally. This study evaluated the performance of five Convolutional Neural Network (CNN) models in classifying the quality and maturity of dragon fruit. The models included MobileNetV3 (small and large), EfficientNetV2-S, NasNet Mobile, and DenseNet-121. Among the evaluated models, NasNet Mobile was the most accurate for maturity grading, achieving an accuracy of 96.64%. On the other hand, EfficientNetV2-S achieved the highest performance for quality classification, with an accuracy of 98.79%. These lightweight CNN models were successfully integrated into a mobile application, PitAIya, designed to classify the quality and maturity of dragon fruit. This research provides a promising direction for future improvements in dragon fruit farming by offering a time-effective and accurate way of assessing quality and maturity.