TY - GEN
T1 - Minor Surface Cracks Detection using SCD11 Convolutional Neural Network
AU - Hussain, Adil
AU - Qureshi, Kashif Naseer
AU - Zaman, Faizan
AU - Aslam, Ayesha
AU - Tariq,
N1 - Publisher Copyright:
© 2024 IEEE.
PY - 2024
Y1 - 2024
N2 - The manual detection of road cracks is a time-consuming process. On the other hand, solutions that are based on deep learning are both speedy and accurate. Recently, several different Convolutional Neural Networks (CNN) based on deep learning have been proposed. However, the performance of the CNN models has varied. The major challenge is the computational resources required to train a pre-trained CNN model; however, a lightweight CNN is more suitable for better training efficiency. In this study work, the SCD11 CNN model is implemented and compared with the pre-trained CNN models, including Inception V2, VGG19, and Xception CNN. The models are trained and tested using the public dataset i.e., the Surface Cracks Dataset. The dataset is divided into training, validation and test sets. The SCD11 CNN along with the pre-trained CNN models are trained and validated and then tested using the splitting of the public dataset. Furthermore, the model's performance evaluation is performed by using a private dataset. The results show that the SCD11 CNN performs better than the pre-trained CNN models for both the public and private datasets.
AB - The manual detection of road cracks is a time-consuming process. On the other hand, solutions that are based on deep learning are both speedy and accurate. Recently, several different Convolutional Neural Networks (CNN) based on deep learning have been proposed. However, the performance of the CNN models has varied. The major challenge is the computational resources required to train a pre-trained CNN model; however, a lightweight CNN is more suitable for better training efficiency. In this study work, the SCD11 CNN model is implemented and compared with the pre-trained CNN models, including Inception V2, VGG19, and Xception CNN. The models are trained and tested using the public dataset i.e., the Surface Cracks Dataset. The dataset is divided into training, validation and test sets. The SCD11 CNN along with the pre-trained CNN models are trained and validated and then tested using the splitting of the public dataset. Furthermore, the model's performance evaluation is performed by using a private dataset. The results show that the SCD11 CNN performs better than the pre-trained CNN models for both the public and private datasets.
KW - Cracks Detection
KW - Deep Learning
KW - Minor Cracks
KW - Road Cracks
KW - SCD11 CNN
UR - http://www.scopus.com/inward/record.url?scp=85203686864&partnerID=8YFLogxK
U2 - 10.1109/eSmarTA62850.2024.10638912
DO - 10.1109/eSmarTA62850.2024.10638912
M3 - Conference contribution
AN - SCOPUS:85203686864
T3 - 4th International Conference on Emerging Smart Technologies and Applications, eSmarTA 2024
BT - 4th International Conference on Emerging Smart Technologies and Applications, eSmarTA 2024
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 4th International Conference on Emerging Smart Technologies and Applications, eSmarTA 2024
Y2 - 6 August 2024 through 7 August 2024
ER -