| Popis: |
Deep learning methods, in their latest progress, enable key steps in object detection and recognition to be performed with convolutional neural networks (CNN). In this article, we apply Yolo (You Only Look Once) algorithm on areal imagery for spatiotemporal tracking of vehicles, for GIS applications. Data augmentation technics are used to increase the number of training and test data. After augmenting the aerial images of vehicles and their corresponding labels, Yolo is trained in iterations until reaching an acceptable precision. Then the trained algorithm is applied on frames of an aerial video of a parking lot, whose outputs (bounding boxes) are used in the LinktheDot algorithm in order to track vehicles in space and time. |
