IHEANYICHUKWU, Chigozirim Goodnews  <br/><br/>
NEURAL NETWORK CLASSIFICATION OF ASPHALT PAVEMENT DEFECTS/ 
Chigozirim Goodnews IHEANYICHUKWU; Supervisors: Mohammed Ali MOSABERPANAH,  Umar Ozgunalp 
 -  2020 
 - VII, 63 sheets;  tables, figures, illustrations,  30.5 cm  CD. 
<br/><br/>Thesis (MSc) - CYPRUS INTERNATIONAL UNIVERSITY INSTITUTE OF GRADUATE STUDIES AND RESEARCH Civil Engineering Department 
<br/><br/>Includes bibliography sheets 57-63 
<br/><br/> ABSTRACT<br/>Asphalt pavements are prone to deterioration from the moment they are laid, these deteriorations are in form of cracks. The most common type of these cracks are linear cracks, transverse cracks, alligator cracks and potholes. Asphalt pavement conditions are rated with the Pavement Condition Index which involves manually identifying cracks and their severity. This manual method of identifying cracks is time consuming while lots are spent of financing the endeavor. This thesis aims to classify pavement cracks using a pretrained neural network. Automatic pavement classification can automate the tasks involved in calculating the Pavement Condition Index while cutting down on the financial aspects. There is still work to be done with data collection and data availability but the results show promise. The network has a training accuracy of 92.86% but struggles to classify images into their correct classes with a precision of 56.57%.<br/>Keywords: asphalt, pavement classification, neural networks, machine learning, Alexnet 
<br/><br/><label>Subjects--Topical Terms: </label><br/>Asphalt <br/>Civil engineering