The Application of Convolution Neural Network for Coconut Maturity Classification

Coconut's sweetness consistency has a close connection with the quality of coconut's byproducts. And this is due to multiple factors. One of which is the maturity of coconut. Maturity classification is, therefore, the main concern for both consumers and manufacturers. However, currently the maturity classification process is mainly based on individual experience and skills developed throughout the time in the industry. Our research proposes an automatic maturity classification with a deep learning which could minimize the maturity classification time cost. The fully trained CNN structure with fine-tuning of transfer learning from well-known models were analyzed and compared. The experimental results showed that our proposed CNN structures achieved the best validation accuracy by 4-fold cross validation results with 97.27% with Binarized Saturation images, and 89.42% with HSV images.