Abstract: The recent advancement in artificial intelligence approach or deep learning techniques explored the ways to facilitate automation in various sectors. The application of deep learning with the computer vision field has resulted in the realization of intelligent systems. World Health Organization (WHO) estimates the road traffic death in India to be 22.6 per 1,00,000 population. Several factors like lousy drivers, appalling road conditions, ignorance of the traffic rules, inability to understand the present road situation and making correct decisions instantly, may contribute to the road crashes and the eventual deaths. Thus an Intelligent Vehicle System has become very important in today's world which will provide an aid to the driver for dealing with road classification, identifying complex road situations and alert him or her beforehand about the probable
crash.
The technique to be used for vehicle detection is Mask -RCNN. The Mask R-CNN model predicts the class label, bounding box, and mask for the objects in an image. It will be accomplished using an online GPU and cloud services provided by Google Colab by using Tensorflow and Keras framework.
The project should successfully detect each car in the image and mark them with independent masks and a bounding box of just the accurate size to fit in the segmented object. We will be able to observe that almost all the vehicles are recognized by the trained model. The model performs satisfactorily for occluded and small sized objects as well.
Keywords: Computer Vision, Image Classification, Scene Detection, Machine Learning, Convolution Neural Network, Places Dataset.
|
DOI:
10.17148/IJARCCE.2022.117116
