| Abstract: |
Fruit ripeness classification is a critical task in the agricultural supply chain, directly influencing harvesting decisions, sorting efficiency, storage planning, and consumer satisfaction. Traditional ripeness assessment methods rely heavily on manual inspection, which is subjective, labor-intensive, time-consuming, and prone to human error and inconsistency across large-scale operations. With the rapid advancement of deep learning and computer vision technologies, automated systems capable of classifying fruit ripeness with high accuracy have emerged as a promising alternative. This review paper presents a comprehensive meta-analysis of past research works employing deep learning techniques including Convolutional Neural Networks (CNNs), transfer learning models, hybrid architectures, and vision transformers for automated fruit ripeness classification [1]-[5]. The paper systematically surveys datasets, preprocessing techniques, feature extraction strategies, and classification algorithms used across various studies involving fruits such as tomatoes, bananas, mangoes, apples, and papayas [6]-[10]. A critical analysis is conducted to identify strengths, limitations, accuracy benchmarks, and computational trade-offs of existing approaches. The methodology adopted for this review involves systematic literature collection, categorization based on model architecture, and comparative evaluation of reported performance metrics. Key challenges such as limited dataset diversity, lighting variability, background noise, and generalization across fruit varieties are discussed in detail. The review concludes that while CNN-based and transfer learning models have achieved substantial success, there remains significant scope for improvement through lightweight architectures suitable for real-time embedded deployment, multimodal sensor fusion, and larger standardized datasets. This paper aims to serve as a foundational reference for researchers pursuing further innovation in inte |
