Abstract:
The timber sector is essential to the economy; however, it continues to depend on manual and often
inaccurate techniques for identifying defects, grading, and cutting. As demand increases and
sustainability issues become more pressing, there is an urgent requirement for smart, automated
solutions. This project aims to tackle these inefficiencies by utilizing artificial intelligence to
improve precision and decision-making in the inspection and processing of timber.
To tackle this issue, TimberTrack was created a web-based application that leverages deep learning
and image processing techniques for the detection of timber defects, grading, and optimization of
cutting processes. The system utilizes YOLOv5 for real-time identification of defects across five
categories (Crack, Hole, Knot, Combined, Good), relying on annotated datasets. A convolutional
neural network (CNN) classifier assesses timber quality, while a tailored rule-based logic suggests
the most efficient cutting paths to reduce waste. The backend is developed using Python, and the
frontend is constructed with ReactJS, enabling users to upload images, view outcomes, and
produce reports.
The performance of the model was assessed through metrics such as Precision, Recall, F1-Score,
and mean Average Precision (mAP). The detection model attained a mAP of 91.3% at an
Intersection over Union (IoU) of 0.5, while the classifier achieved an accuracy of 94.7%.
TimberTrack exhibits significant promise as a dependable, AI-powered solution for improving
quality control within the timber sector. Future enhancements may include expanding the dataset,
incorporating real-time video capabilities, and integrating explainable AI features.
