Deepcrack – A Deep Learning Approach for Image Based Crack Prediction
R. Venkateshwarlu
Associate Professor
Dept. of Computer Science and Engineering
Jyothishmathi Institute of Technology and Science
(JNTUH)
Karimnagar, Telanagana, India
rudrarapu.venkateshwarlu@jits.ac.in
Rimsha Mohammadi
UG Student
Dept. of Computer Science and Engineering
Jyothishmathi Institute of Technology and Science
(JNTUH)
Karimnagar, Telanagana, India rims.md22@gmail.com
Erram Akshaya Reddy
UG Student
Dept. of Computer Science and Engineering
Jyothishmathi Institute of Technology and Science
(JNTUH)
Karimnagar, Telanagana, India akshayareddyerram@gmail.com
Sunkari Rahul
UG Student
Dept. of Computer Science and Engineering
Jyothishmathi Institute of Technology and Science
(JNTUH)
Karimnagar, Telanagana, India rahulsunkari887@gmail.com.in
Edla Bhavani
UG Student
Dept. of Computer Science and Engineering
Jyothishmathi Institute of Technology and Science
(JNTUH)
Karimnagar, Telanagana, India
bhavaniedla7@gmail.com
R.Shashank
UG Student
Dept. of Computer Science and Engineering
Jyothishmathi Institute of Technology and Science
(JNTUH)
Karimnagar, Telanagana, India
radandishashank@gmail.com
Abstract— Crack detection is critical for the safety and longevity of civil infrastructures like roads, bridges, and buildings. However, the conventional approach for crack detection involves manual visual inspection, which is not only time-consuming but also vulnerable to errors. To overcome these challenges, the present project proposes an automated system for crack detection, severity analysis, and cause prediction using a deep learning approach, named "DeepCrack: A Deep Learning Approach for Image-Based Crack Prediction." The proposed system is implemented using a full-stack approach with a React-based frontend and a Flask-based backend using the Python programming language. For image preprocessing, the system utilizes the OpenCV and Pillow libraries. For crack detection, the system utilizes the U-Net++ model implemented using the PyTorch library with various image augmentation techniques like rotation, flipping, scaling, and brightness adjustment. After crack detection, the system predicts the severity of the crack based on the crack area and distribution. Additionally, the system predicts the cause of the crack, including fatigue, moisture, thermal expansion, corrosion, and overloading.
Keywords— Deep Learning, Crack Detection, U-Net++, Crack Segmentation, Computer Vision, Structural Health Monitoring, Data Augmentation, Severity Analysis, Image Processing, Infrastructure Inspection.
