This repository contains the scripts used in the workflow to create The HAGDAVS dataset, a dataset to detect and segment vehicles (motorcycles, cars, and ghsots of vehicles). The dataset is obtained from drone orthomosaics and it will be publicly available.
The following figure is an example of resulting images and bounding boxes of the HAGDAVS dataset. It has three classes: mortorcycles, cars, and ghosts, with codes 0 to 3, and the background as a fourth class.
Images are obtained after tessellation of RG-DSM and Multi-class Mask Image, as shown in following figure.