Artistic Style Transfer of Satellite Imagery: Sentinel-2 & RASAT
When I first see the Copernicus post about converting its Sentinel-1 picture into Van Gogh’s the Starry Night painting, I asked myself “Why not make the same for satellite imagery?”. The Earth itself is the finest art if you look from space. Even Google Earth provides a few thousand views of the Earth: Earth View.
The application is based on Leon A. Gatys’ paper, A Neural Algorithm of Artistic Style. There is a thorough explanation at several on the web.
An artistic style is transferred to other imagery by using deep neural networks.
Later, I was inspired by ESA’s tweets on converting the artist’s drawing of Sentinel-1 to Van Gogh’s famous the starry night painting.
I also converted TÜBİTAK UZAY’s RASAT satellite imagery with some flavor of Turkish Culture: Turkish carpets, Turkish water marbling, and Turkish Tile Art.
This tile art suited to İstanbul as it contains blue and red colors already. I may think of printing this one as a poster. Well, good idea.
Tulip was introduced by Turks to the world. It is also Turkey’s official symbol.
I loved this red carpet, it is a reflection of İstanbul’s cosmopolite demography.
Van Gogh’s The Starry Night painting.
Moreover, I obtained Turkey’s July 2019 mosaic of Sentinel-2 by using Google Earth Engine and applied some famous paintings on it.
I also converted Sentinel-2 imagery of İstanbul to Starry Night (again). Did you notice seams on the imagery. Sentinel-2 camera is a combination of several detectors which are equalized during the image processing chain.
I study Crop Mapping of the Harran Plains in my PhD, it would be nice seeing it in different modalities:
Which styles would be suitable for adding some artistry to satellite imagery? Paintings which contains the Earth’s colors fit nice?
If you are looking for beautiful RASAT imagery see Gezgin Blog.