Innovative Tomato Rot Detection: A Deep Learning Approach using EfficientNet-B0 for Farming Efficiency

Tomato cultivation plays a crucial role in global agriculture, yet post-harvest losses due to rot significantly impact both yield and market value. Traditional methods of detecting rot are manual, subjective, time-consuming, and prone to human error. These limitations result in delayed identification of decayed produce, affecting the entire supply chain from farm to consumer. Furthermore, existing automated approaches often require high computational resources and lack robustness under real-world conditions such as varying lighting, occlusion, and background noise. This paper presents an efficient deep learning-based solution for automated tomato rot detection using the EfficientNet-B0 architecture. EfficientNet-B0 is known for its optimized performance-to-complexity ratio, making well-suited for deployment on edge devices in agricultural settings. The proposed system leverages a curated dataset of diverse tomato images, capturing different stages of rot under natural environmental conditions. The methodology includes robust data preprocessing techniques and transfer learning to fine-tune EfficientNet-B0 on the domain-specific task. The model is trained using binary crossentropy loss and evaluated using key performance metrics such as accuracy, precision, recall, and F1-score. Experimental results demonstrate a classification accuracy with precision and recall values respectively. These results validate the effectiveness of the proposed system in accurately detecting rot while maintaining low computational demands. This work addresses current agricultural challenges by enabling real-time, reliable rot detection in tomato farming, thereby improving crop quality management and reducing post-harvest losses.