Review on Methodologies of Object Detection

Abstract

This project is an application of Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN) in the field of object detection and classification. CNN's are best applicable in image and video recognition. The system proposed in this project involves training the network over images and processing the input video frames for testing. The model will be trained over images of potholes, road signs and pedestrians. The dataset of images for potholes is created, as there is no specific dataset available. The dataset of images for road signs and pedestrians is created by collecting images from various sources and formatting them. The model will be trained over these datasets and tested on a real time video. This is a prototype which can be implemented in automated cars and can be used by car drivers as an Android application, which detects the objects and alerts the user through a voice message.