Resources – Data
It is our intention to collect information about benchmark data that can be used to compare pattern recognition methods on relevant remote sensing tasks. The following list contains items listing task, comment, and source web site.
- Berlin-Urban-Gradient dataset 2009 – An EnMAP Preparatory Flight Campaign
- TorontoCity: Seeing the World with a Million Eyes (coming soon)
- DigitalGlobe SpaceNet
- Inria Aerial Image Labeling Dataset
- UC Merced Land Use Dataset
- Zurich Summer Dataset
- Brazilian Coffee Scenes Dataset
- IEEE GRSS Data Fusion Contests
- Semantic Labeling Benschmark in 2D and 3D: The ISPRS working group III/4 “3D scene analysis” released two new benchmarks:
- Semantic Labeling Benchmark: The ISPRS working group III/4 has released a 2D semantic labeling benchmark. 33 true ortho image and DSM tiles of very high resolution and quality are provided, and in addition classification labels for 16 of those tiles are given as training set. The remaining 17 tiles are the test set.
- Segmentation data-generator and benchmark: Synthetic texture images are generated with groundtruth segmentation. This is possible, e.g. also from ALI hyperspectral satellite images
- More datasets