Paper Title
Image Processing Based Vehicle Detection Using Deep Learning

Abstract
Single Shot Multi-Box Detector (SSD) has a tendency to provide unpleasant results, particularly for tiny targets such as automobiles on high-resolution photos. This is the case despite the fact that SSD has a high level of accuracy and a rapid speed when it comes to object recognition. Detecting automobiles on high-resolution photos is the focus of our study, in which we present a novel convolutional neural network that is built on solid-state drives (SSD). Both the feature fusion module and the detection module have been implemented into the framework that has been proposed there. Feature maps of varying scales are included into a fusion feature for object detection within the feature fusion module. This integration has the potential to significantly increase the accuracy of the detection process. In addition, the batch normalization layer is placed between the detection layers in the detection module. This is done to prevent the network from overfitting and to speed up the training process. Experiments involving ablation give substantial evidence that the structures described above are effective. We are able to attain an average accuracy of 0.904 on the UCAS-High Resolution Aerial Object Detection Dataset with our network, which is 0.094 AP greater than SSD512 but has a performance that is comparable to it. Keywords - Deep Learning, Vehicle Detection, SSD, High Resolution, Convolutional Neural Network