IMPLEMENTASI MODEL CONVOLUTIONAL NEURAL NETWORK (CNN) UNTUK DETEKSI KESEGARAN BUAH PISANG BERDASARKAN CITRA KULIT

Abstrak

The assessment of banana freshness is currently still done manually through visual observation, touch, and smell. This method is subjective and prone to errors in perception between individuals, which can cause losses for farmers, traders, and costumers. Inaccuracies in assessing freshness levels can result in the distribution of substandard fruit, reduced market competitiveness, and waste of resources. To address these issues, this study designed and implemented a banana freshness classification system using a Convolutional Neural Network (CNN) algorithm. The system was develoved in the form of a Python and Flask-based website. Equipped with a Text-to-Speech (TTS) feature to improve accessibility for users with visual impairments. The research stages included problem identification, banana image data collection, image preprocessing (resize, normalization, augmentation), CNN architecture design, model training, implementation, and testing. The dataset consist of 1,664 images classified into two categories: fresh and not fresh. The implementation result show that the system can classify banana freshness in real-time through visual and audio displays. This system has the potentional to improve the efficiency and objectivity of classification, as well as support the digitization of the agricultural sector.