The AeroScapes aerial semantic segmentation benchmark comprises of images captured using a commercial drone from an altitude range of 5 to 50 metres. The dataset provides 3269 720p images and ground-truth masks for 11 classes.
Clone the repository
git clone [email protected]:ishann/aeroscapes.git
Download the data
This results in the following directory
data/ aeroscapes/ JPEGImages/ 3269 RGB images. SegmentationClass/ 3269 ground-truth segmentation masks. Visualizations/ 3269 RGB ground-truth segmentation visualizations. ImageSets/ Training and validation splits for data. aeroscapes.tar.gz Downloaded file (local reference to avoid need for repeated downloads).
If you use AeroScapes in your research, please cite the following:
Ensemble Knowledge Transfer for Semantic Segmentation Ishan Nigam, Chen Huang, Deva Ramanan Proceedings of the 2018 IEEE Winter Conference on Applications of Computer Vision