What is AutoMap and who are the brilliant minds behind it?

AutoMap is a complex workflow reduced to a simple user interface that produces maps based on freely available satellite images. We take care of downloading, processing and merging terabytes of data and classify what matters to you. AutoMap is a product of the Swiss ExoLabs GmbH, a spin-off from the University of Zurich. We are scientists and remote sensing experts.

Who can profit from AutoMap?

Changes in environmental conditions affect most of the world's population. Accordingly, not only policymakers but also industry needs to adapt to these changes and needs to develop strategies to monitor environmental processes. For this it is essential to know where on earth changes occur, what kind of changes happened and to what extent the changes took place. AutoMap provides a unique solution for insurance companies, financial institutions, commodity traders, tourism industries, energy companies and scientific institutions.


How long will a job take to finish?

Obviously it depends on your input parameters. A big area of interest and a long time interval will increase the number of individual images to process. When classifying, every single pixel needs to go through the classification pipeline. In general, this is not an on-demand application.

To give you one isolated example:
Area: ~ 3000 km2
Time interval: 2 Months
Satellites: All (196 scenes)
Mode: Default classification
Runtime: ~ 1 hour
Cost: ~ 1.50 $


A single raw Sentinel-2 scene has almost 2 GB and about 120 million pixels!

My Job is still running, did it freeze?

For every run and machine we create an individual alert system that notifies us about run-time problems. In the rare occasion of a frozen code, the team will reach out to you and take care of a fast solution. In case of a real crash you will be notified per mail as is the case for successfully finished runs, so while the monitoring system tells you that the code is still running and nobody reached out to you, the code is actually still running. The whole workflow is complex and time-intensive, even on optimized cloud machines.

I was notified that an error occurred, what should I do?

First of all, sorry for the inconvenience! In case of a runtime-crash the team will be notified as well by default. We take those things very seriously since it means we are not as brilliant as we thought we were! The email with which you were notified contains the last log message that maybe helps identify the problem right away. The full log file in the download container should reveal everything. If nobody reaches out to you because we concluded an obvious mistake from your side but you're still unsure, don't hesitate to contact us.

Why are results only stored for 48 hours?

Cloud solutions are very cost efficient but still not free. Considering that results are multiple gigabytes big, the 48 hour policy helps us to reduce costs for you and us!


What is the output of an AutoMap run? How can I interpret the results?

The output you receive is a link to a zip-file on an Amazon AWS S3 bucket that is valid for 48 hours. If you download and decompress the folder, it contains all your results:

  1. AutoMap.log
  2. It contains the log connected to the algorithm itself. In there you find a complete dictionary of all used images, their cloudcover and geometry. In case of a classification it also holds information on the classifier, namely the hyperparameter values, the Cohen's Kappa and sample numbers.

  3. Raster Files
  4. Depending on the operating mode you receive the different output rasters per selected satellite. A file named only after the respective satellite holds the reduced results over the time dimension. If the file is ending with the keywords With VIs, it further contains Vegetation Indices used for classifications. FInally, a file ending with the keyword Classification holds the classification result (1), it's certainty per pixel (2) and cloud mask (3).

  5. ReadMe
  6. For every selected satellite, the respective Readme contains metadata information about all used single scenes and the order of bands in the resulting raster file.

  7. Slices
  8. Under the hood we apply a divide-and-conquer approach, slicing the complete area of interest in smaller sub-tiles. This vector holds these tiles and could help identify problems.

  9. Crosstables
  10. The classification crosstables per satellite are stored in csv files that hold the responding Cohen's Kappa in the filename.

What do the values of the cloud mask mean?

The cloudmask differentiates no clouds (1), cloud (0) and out of bounds (-1).

I used the default classification input (MCD12Q1 v006), what is it exactly?

The default classification input is based on MODIS data that is aggregated to annual global land cover classification maps from 2001 - 2018 where the closest respective year is selected. For more information, please refer to:

General Information
User Guide (Class Labels)


How does the pricing work?

We offer a free and a premium pricing model. For all registered users a limited pool of machines is made available for free use. When this pool is occupied your job will not be submitted. For the premium model, the approach is simple. When started, we record the current hourly run-time cost of the launched cloud machine. Considering expenses for storage and transfer and adding our win margin, we simply double that price.

At this moment, the unit price is 1.552 $ per hour and job.

How can I change my pricing model?

If you would like to become a premium user, please contact us!

My run failed, do I need to pay anyways?

No! We will not charge you for failed runs. A run is considered to have failed when the database records a negative status.


Where can I get more support?

Don't hesitate to contact us via Email.

I want to do something different with this algorithm or even change the code, is that possible?

Don't hesitate to contact us via Email.