Wood Classification with Transfer Learning Method and Bottleneck Features

Abstract

Wood is one of the raw materials that is important for helping human needs such as home appliance like chair, table, and the other things made from wood. Wood industries require wood for produce their product that made from wood. There are several classes of wood must be classified before entering the process stage in wood industries. The class of wood is indicated the quality of the wood itself. So, wood industries must classify carefully for increase the productivity. Wood industries usually classify the wood with manual method such as classifying with eyesight of human however it just produces about 55% of correct classification rate. Convolutional Neural Network (CNN) is a development from Artificial Neural Network (ANN) to classify the image, image segmentation, and object recognition with high accuracy and high performance. Wood classification is one of the texture classifications because the wood can be classified depend on texture of the wood fiber itself. So, Convolutional Neural Network (CNN) comes to solve this problem. Deep Convolutional Neural Network (D-CNN) is developed to improve the accuracy and performance from many deep layers of Convolutional Neural Network (CNN) model and is designed for classify image with small dataset. This paper investigates the usage of Deep Convolutional Neural Network (D-CNN) with transfer learning method and bottleneck features for wood classification with a small dataset. There are five different classes of wood classifications in this paper such as class I, class II, class III, class IV, and class V. The best quality of this wood is class I and the worst quality in this wood is class V. Transfer learning method and bottleneck features can increase the accuracy of Deep Convolutional Neural Network (D-CNN). In this paper, the experiment achieves 95.69% of accuracy with transfer learning method and bottleneck features.