Progressive Image Resizing for wood Species Classification from Macroscopic Images

Mazen, Fatma Mazen Ali; Nashat, Ahmed Aly; Ibrahim, Marian Mamdouh

doi:10.21608/fuje.2024.276630.1070

Progressive Image Resizing for wood Species Classification from Macroscopic Images

Document Type : Original Article

Authors

¹ Fayoum University - Faculty of Eng. - Dept. of Electronics and Communication Eng. Fayoum 63514

² Fayoum University Faculty of Engineering Electronics and Communication Engineering Dept.

³ Electrical Engineering Dept., Ahram Canadian University, 6 October City, Giza, Egypt

10.21608/fuje.2024.276630.1070

Abstract

Wood is an important raw material used in various activities, such as building, furniture, and fuel. The timber industry is significant in many countries and has a significant financial impact. There are diverse categories of wood, each with its unique properties, and experts typically perform wood species identification through visual inspection, which is a tedious and time-consuming process. To eliminate the need for manual detection, a deep learning-based wood Classifica-tion System was proposed in this paper. The system uses a transfer learn-ing-based convolutional neural network model that handles feature extraction. Compared to other transfer learning models such as VGG16, ResNet50, and DenseNet201, the proposed EfficientNetB7 model achieved a high validation accuracy of 99.824%, which suggests that it can be used to aid unskilled agents in wood categorization. This new strategy can save time and effort in the identi-fication of wood species, making it an efficient method for the timber industry.

Keywords