Classification of Pineapple Fiber Woven Fabrics Based on Convolutional Neural Network

Abstract

In this research, we propose a method for classifying 4 types of pineapple fiber woven fabrics which are pure pineapple fiber woven fabrics, manual-connected pure pineapple fiber woven fabrics, machine-connected pure pineapple fiber woven fabrics, and blended pineapple fiber-cotton textile fabric. Each type can refer to the quality and price of these fabrics. The proposed technique is based on transfer learning of the Convolutional Neural Network (CNN). Transfer learning can be separated into 3 steps which are modifying the pretrained network, model training, and assessing the model. The Resnet50 was selected to be our transfer learning network. Finally, the experimental result shows that the classification performance in terms of class accuracy is 95.83 % on average.