Counting Cars from Aerial Videos Using Deep Learning

Counting vehicles in heavy traffic areas is not an easy task, even for a trained human. Recent advances in real-time object detection using convolutional neural network can jointly detect and identify objects. We propose a straightforward application of using real-time object detection algorithm to count the number of vehicles in high traffic areas. The user defines marks on the road where the proposal counts the number of vehicles crossing the marks. The proposal achieves 87.24% average correct counting in a video sequence with real-time performance.