Skip to content

Cocktail - resource constrained, collaborative geospatial analysis and visualization.


Notifications You must be signed in to change notification settings


Folders and files

Last commit message
Last commit date

Latest commit


Repository files navigation


Resource constrained, collaborative satellite asset collection and analysis

cocktail tip and cue

Cocktail is designed to facilitate resource constrained collaborative inquiry on optical satellite assets. Resource constrained collaboration is supported in several different ways. Cocktail will vary parameters of a classifier in batch mode and store the classifier performances for each setting so you can collaboratively can find the best configuration. Cocktail keeps track of the large collection of parameters across the individual processing modules in a single file such that remote collaborators can reliably replicate a workflow. Moreover, Cocktail can be linked with an external cloud storage provider to move assets out of the compute environment for low-cost storage of large GIS files.

Cocktail is designed as a compendium to other GIS investigation environments, combines GDAL. OTB and GQIS elements, and is intened to be deployed remotely. The combination requires some attention. The Orfeo machine learning library and the translator library for raster and vector geospatial data formats GDAL do not play well with the open source geographic information system QGIS outside of a resource intensive GUI environment. Cocktail allows you to setup a low-cost virtual computer to apply several different classifiers including vector support machines, random forests, neural networks and various bandmath operations across GDAL, OTB and QGIS.

Private sector, high-resolution, daily-updated satellite assets have become a significant resource for remote sensing operations. A case in point is PlanetScope (PS), currently operating the largest collection of small Earth-imaging satellites. Cocktail facilitates the combination of free and commercial satellite assets to monitor an area of interest with a low cost system, and then switches to a high resolution asset only if a condition of concern or interest has been detected and not adequately understood in the first data set. Not unlike the tip and cue concept in which one monitors an area of interest with one sensor and then ‘tips’ another complementary sensor platform to acquire another image over the same area, Cocktail tips and cues based on economic constraints.

Check the Install+Use document for details and examples.

Here is a brief description of the steps required to get started:

1 - VM
Create a virtual computer. Choose Ubuntu 18.04 LTS, with at least 16GB of RAM. If you intend on working with Sentinel data from ESA, it can be beneficial to build a VM close to ESA in Frankfurt.

2- Get the repo files from this Github site.

git clone

You will have all the files in the correct directory structure on your computer after the download.

3 - Install QGIS.
Go to the setup folder. Make the file executable.

chmod +x

Run that file.


4 - Install OTB in a virtual environment with conda.
Go to the setup folder. Make the file executable.

chmod +x

Run that file.


5 - Create a conda environment with defined dependencies.
(OTB 7.2 requires python 3.7, for example.)

conda env create -f environment.yml

6 - Customize
Add other libraries to the OTB environment as needed. Enable the OTB environment

conda activate OTB

And install geojson and geopandas. The geopandas install may take a bit of time.

conda install -c conda-forge pillow
conda install -c conda-forge geojson
conda install -c conda-forge geopandas
conda install -c conda-forge sentinelsat

7 – Setup the directory structure
The directories code, data, results should be ready made in your repository directory. Adjust the paths to reflect your current installation. From the setup folder run the adjust_datapaths script:


Follow the prompt. All paths will be adjusted to your current installation and the settings.txt file will be updated accordingly. Put analysis-ready raster and vector files (including shapefiles) into the data/collection directory. Use the settings.txt file (in the data directory) to set your file names, process preferences and classification parameters. The content of this file is parsed and passed to the classification steps.

Change the fake username and passwords in the auth folder according to your needs. PC.txt is for pCloud, sent.txt for sentinel and planet.txt for Planet Labs. Put data from sentinel and planet into the collection directory. Add .geojson references files (for sentinel and planet downloads) and colormaps as needed to the data directory.

8 - Test
a) Test QGIS - (test scripts are in the code directory)

Change to the code directory and deactivate the conda environment

conda deactivate

Run the qgis test in the base environment


b) Test OTB

Activate the OTB environment

conda activate OTB

Run the otb test


c) Update settings

Update the settings and file parameters corresponding to paths on your VM.

nano settings.txt

Then verify that all is ok


You can now use QGIS in the base installation and enable the conda environment to access OTB functionality. You can also move across environments to access libraries from either environment with python scripts as outlined below (see

echo "Starting OTB-QGIS pipeline...\n\n" 
conda run -n OTB python3 /home/code/ 
python3 /home/code/ 
conda run -n OTB python3 /home/code/ 
python3 /home/code/ 

Since the directory structure you setup will be identical in both the QGIS and OTB environments, intermediate filed produced will be available to processes running in either environment, allowing for data to be shared. All settings across the script modules are stored in the 'settings.txt' resource file and imported to the individual modules.

Details on how to collect data from the European Space Agency's Sentinel program, how to make use of tip and cue, and how to perform various band-math operations and image classification with SVM, RF and NN classifiers is described in the Install+Use .pdf document.

The collection directory contains some sample Sentinel2 data. Here are links to three high resolution 8-band (2022) and 4-band PlanetLab images (geoTIFF format, 500MB / 300MB each):

Central Bali, June 2020
Central Bali, May 2021
Central Bali, May 2022

Cocktail has been used to create research artifacts presented at the following venues:



Cocktail - resource constrained, collaborative geospatial analysis and visualization.








No releases published


No packages published