Paper Title
Pedestrian Detection In Autonomous Driving Application Using Convolutional Neural Network

Abstract
Pedestrian detection is of high importance to autonomous driving applications. Methods based on Neural Network have shown significant improvements in detection rate, which makes them suitable for this application in which reducing the False Discovery Rate is very important. Convolutional neural network (CNN) has achieved great success in the field of computer vision. CNN takes input data only as image in fixed size and this arises problem in scaling. Hence this paper discusses a filter based feature extraction. Henceforth, the object identification is done without any size constraint. Pedestrian detection also faces the challenges of background clutter and large variations in pedestrian appearance due to pose and changes in viewpoint etc. One of the key contributions is also towards this issue by training the network accordingly. This paper ultimately focuses on reducing the false discovery rate and increasing the accuracy of the detection method. The precision predictive value obtained is 51.46% with a false discovery rate of 48.54% using the benchmark data. The FMeasure value is 65.35%. The number of iterations to minimize the error was achieved to be 1100 epoch and the classification rate of the input data as objects and background is 97.40%. Keywords— Convolutional Neural Network, Miss rate, False discovery rate, F-Measure.