Obstacle Detection System in Dashcams using Convolutional Neural Networks

Obstacle detection is an essential safety feature in all modern cars. Dashcams can be used to record traffic footage. The images obtained from dashcam have to be analyzed to detect the obstacles. Obstacles in the images need to be classified based on the ir distinctive properties. For a classification task, the object’s feature has to be detected. It is not reliable to hard code to detect all features of any object as it will reduce the accuracy of prediction. A Neural Network is a better approach where it will determine the filters needed to classify the object into its respective class. Hence, a class of deep, feedforward neural networks called convolutional neural networks has been used to analyze the imagery. In this paper, based on convolutional neural network, an efficient and accurate system to identify obstacles using the dash cam footage is being devised.