RCS-Image Dataset

A synthetic dataset for object detection, semantic segmentation and depth recognition was created using images from a virtual environment in Unreal Engine. This dataset that could be used to retrain neural networks on virtual aerial images, for object detection, segmentation and depth planning. There is ground truth for full pixel level semantic segmentation and object detction by way of bounding box coordinates in .exif files. The main  labels present in the dataset are train truck and gas cylinders  The size of the data is currently more than 10GB, with 1000 aerial images, from 100 different waypoints at 10 different heights. For each raw image, there is also an associated depth image, a segmented image, object ground truths with bounding box, and the metadata .exif file. If you found this dataset useful for your reserach, please cite as, @article{smyth2018virtual, title={A Virtual Environment with Multi-Robot Navigation, Analytics, and Decision Support for Critical Incident Investigation}, author={Smyth, David L and Fennell, James and Abinesh, Sai and Karimi, Nazli B and Glavin, Frank G and Ullah, Ihsan and Drury, Brett and Madden, Michael G}, journal={arXiv preprint arXiv:1806.04497}, year={2018} }