I'm working on the algorithm in this paper and I've modified it a bit. Now in order to quantify the accuracy of my algorithm with this algorithm, I need ground truth data of SanFrancisco.
Which has at least five classes:
- buildings (if detailed, it'll be better)
- sea (if detailed better)
- fir trees of the vegetated area
- spruce trees of the vegetated area
- pine trees of the vegetated areas
(Generally, it would be better to have a ground truth as detailed as possible).
Do you know such ground truth or any useful rasterized map to be able to be able to make a ground truth with optical imagery?
Maybe this question is a better fit for opendata.