An open source GAN framework
Sobolt’s GAN framework is openly available on Github. This neural network has been used to train the in1 super resolution model for Sentinel-2 seen below. This service was developed in collaboration with the Phi-Lab of ESA in 2020. Use it to train your own generative image models. For further inquiries please contact us on info@sobolt.com.
Sentinel-2 Super Resolution
Witness the super resolution achieved with in1 - trained with our framework. Below you can find a selection of use cases, from agriculture to urban development. Each linked file includes both the super resolved data and the original Sentinel-2 product. Download for free and see for yourself the enhanced value in1 provides.
in1
original
Asset tracking
Use in1 for improved shipment tracking: uncover the bustling traffic of the Rotterdam harbor for March 21st, April 5th and 20th 2020.
in1
original
Urban development
Previously unseen changes in urban development? Get the hawk’s eye view of Munich’s yearly growth from October 14th 2018, September 29th 2019, September 13rd 2020 and September 3rd 2021.
in1
original
Forestry
in1 offers a unique perspective for following one of Europe’s last primary forests, Białowieża, transitioning from Spring to Summer through March 25th, August 17th and September 21st 2020.
in1
original
Agriculture
in1 provides unprecedented insights into the evolution of crops across seasons. Explore the Californian Central Valley raster from May 1st, June 2nd, August 9th and September 3rd 2020, for a closer look at harvest monitoring.