Mango Fruit Disease Detection by Integrating MobileNetV2 and Long Short-Term Memory

Mango, a prized fruit in tropical regions, ranks as the third most traded tropical fruit globally, following pineapple and avocado. Despite its economic significance, the quality of mangoes is often compromised by various factors, including diseases. This paper introduces an innovative approach to disease detection in mangoes, departing from traditional Convolutional Neural Networks models. Instead, we propose a hybrid model that incorporates MobileNetV2 for feature extraction, LSTM for capturing feature relationships, and dense layers for accurate disease classification. The model is trained on the MangoFruit-DDS dataset, comprising over 1700 images. To optimize training, the dataset undergoes augmentation and segmentation processes, enhancing the model’s ability to recognize and classify diseases accurately. This novel combination of MobileNetV2, LSTM, and dense layers enables the model to discern intricate patterns and dependencies within the mango data. Our goal is to achieve a higher level of accuracy in disease detection, contributing to improved mango production quality. The proposed hybrid model not only addresses the challenges posed by mango diseases but also underscores the potential of advanced models in agricultural applications. By leveraging the MangoFruitDDS dataset and innovative model architecture, this research aims to advance disease detection capabilities, fostering healthier and more lucrative mango harvests.