A Deep Learning Approach for Real Time Road Damage Detection via Image Processing

A Deep Learning Approach for Real Time Road Damage Detection via Image Processing

1D.Vivek, 2G.Supriya, 3G. Sri Chaitanya, 4 Priyanka Dash

1,2,3Student,4Assistant Professor

1,2,3,4Department of Computer Science and Engineering

1,2,3,4Guru Nanak Institutions Technical Campus, Hyderabad, Telangana, India

ABSTRACT :

Due to the use of subpar building materials in the construction of road drainage systems, traffic accident identification and prevention are becoming increasingly tough and challenging issues in India. The aforementioned issues lead to early road deterioration and potholes, which result in accidents. An estimated 4,64,910 incidents occur in India annually, according to a report published by the Ministry of Road Transport and Highways' transport research wing in New Delhi. This study suggested a deep learning-based approach that uses photos to identify potholes early, lowering the risk of an accident. Inception-V2, Faster Region-based Convolutional Neural Network (F-RCNN), and Transfer Learning serve as the fundamental foundations for this model.There are fewer pothole recognition models that just use machine learning methods to detect potholes, but several models combine machine learning methods with accelerometer data (without the need for images).  The study's results show that our proposed model outperforms other pothole detecting techniques currently in use.