An Accurate Car Counting in Aerial Images Based on Convolutional Neural Networks

This paper proposes a simple and effective single-shot detector model to detect andcount cars in aerial images. The proposed model, called heatmap learner convolutionalneural network (HLCNN), is used to predict the heatmap of target car instances. Inorder to learn the heatmap of the target cars, we have improved CNN architecture byadding three convolutional layers as adaptation layers instead of fully connectedlayers. The VGG-16 has been used as a backbone convolutional neural network in theproposed model. The proposed method successfully determines the number of carsand precisely detects the center of target cars. Experiments on the two different cardatasets (PUCPR+ and CARPK) show the state-of-the-art counting and localizingperformance of the proposed method in comparison with existing methods. Also,experiments have been conducted to examine the effect of data augmentation andbatch normalization on the success of the proposed method. The code and data will bemade available here [https://www.github.com/ekilic/Heatmap-Learner-CNN-for-Object-Counting].