Roads have an important role in modern societies allowing a comfortable, fast and cheap way to travel from one place to another. As a result, many vehicles use them every day, causing a continuous degradation of the road surface. If an appropriate maintenance policy is not applied, the quality of the road surface degrades,compromising road security. Hence with this knowledge ,my aim to make a damage detection system that detects damages on a road and also classify them into eight uniquely known damages using data set of 9053 images. Successful application of deep learning techniques for damage detection rely on discriminative and representative deep features.
So the idea is to develop a crack detection method in which the discriminative
and differentiative features are learned directly from raw image patches using the Convolutional Network. Here uses SSD(Single Shot detector) which detects objects in a single forward pass of network.
As the diagram above shown, SSD’s architecture builds on the venerable VGG-16 architecture, but discards the fully connected layers. The reason VGG-16 was used as the base network is because of its strong performance in high quality image classification tasks and its popularity for problems where transfer learning helps in improving results.
Here uses the state-of-the-art object detection method(SSD), and compute
the accuracy and run time speed on a GPU server. At the end the system shows that the type of damage can be distinguished into eight types with acceptable accuracy by applying the proposed object detection method.
Author: Riya P D, Member, AI Club VAST