Julien Osman

Ingénieur développement logiciels. Docteur en télédétection.

Career


Lisez-moi en français

Since April 2020

Technical manager at CS GROUP - France

Development of the remote sensing library Orfeo Toolbox.

July 2015 - March 2020

Software Engineer at Thales Services SAS

2020

Development of simulation tools for the MicroCarb project

Setting up of a protocol for the generation of simulation images for the MicroCarb project. Development of tools (Python) to easily run the simulations on the CNES' calculation cluster.

2019

Technical support for the Image Quality service at CNES

The Image Quality as CNES develops multiples algorithms to assess the FTM of satellite images. Creation of a GUI (Qt PyQt) for those algorithms. Setting up of virtual machine to make the tools easy to use on various environments. Creating tests to assess the impact of vibration on a test bench.Generation of images simulations for the validation of the Darwin-CU chain.

Technical support for the atmospheric correction of Sentinel-2 products

In the framework of a competition organized by the NASA and the l'ESA, the CNES uses it's atmospheric correction and cloud detection algorithm MAJA over big volume of data. Development of tools for the preparation of the input data, launching the processing on the HPC, and the delivery of the output data. Python development using linux and Pycharm. Running the processing on a HPC using PBS. Configuration management with Git et Gitlab.

Development of a proof of concept using images localization

Industrialization of LibGeo, a C++ library for satellite images localization. This library allows one to compute the localization (latitude/longitude) of a pixel according to it's position in the image, and inversely. Use of this library for a proof of concept for the generation of DEM from a couple of images.

2018

Mission at IRT Saint-Exupéry

Work on Deep Learning applyed to satellite image analyse (target detection, classification, change detection). Use of active learning techniques to compensate the lack of training data. Development of an active learning application allowing the training of deep neural networks over the cloud, and the labeling of wisely selected samples by presenting them to an operator to increase the training set. Python development using linux and Pycharm. Use of the Tensorflow and Keras library. Processing distributed on the Google Cloud Platform. Configuration management with Git and Phabricator.

Study: Large Scale Object Based Image Analysis

Scaling object oriented, segmentation algorithms base on adjacency graph. Use of the C++ library Orfeo ToolBox. Processes distributed on a HPC, with synchronisation between the nodes using MPI. This work is available on GitHub.

2016-2017

**Design and development of a calibration workshop.

Implementation using Python of various approaches for in-flight satellite calibration. Principal library used: Numpy, Pandas et GDAL. DevOps. Configuration management with Git, Gitlab and RedMine.

2015-2016

Development of an application for satellite image processing used in image quality centers

Image processing algorithms implementation (for image quality assessment). setting up of an orchestrator for on-ground processing. Development of an on-bord configuration files manager. Python development using linux and Pycharm. Use of the SAG framwork (Structure d’Accueil Générique) with Isatis-ng. Use of the GDAL, Numpy, MatPlotLib and gnuPlot libraries. Use of the Phoebus-Lite orchestrator. Configuration management with SVN, Jira and Maven.

April - July 2015

Research Engineer at Centre d'Étude Spatiale de la BIOsphère (CESBIO)

Development of a data access object (DAO) for the DART model using Python.

February - April 2015

Research Engineer at Centre d'Étude Spatiale de la BIOsphère (CESBIO)

Integrating the tools used during my PhD (Markov Logic Network) into the toolbox of the laboratory.

November 2011 - February 2015

PhD Candidate at Centre d'Étude Spatiale de la BIOsphère (CESBIO)

Evaluation of the influence of expert knowledge and prior information on the generation of land-cover maps. Development of test algorithms with a C++ library. Use of Markov Logic Networks and Support Vector Machine. For more information, visite the dedicated page.

September 2012 - March 2013

Secondment at Instituto National de Pesquisas Espaciais (INPE - Brazilian National Space Agency)

Evaluation of the influence of expert knowledge and prior information on the generation of a probability map to monitor the areas with a high likelihood of being deforested.

February - November 2011

Research Engineer at Centre d'Étude Spatiale de la BIOsphère (CESBIO)

Fusion of Sentinel-2 and Proba-V/Sentinel-3 Images for Multi-Temporal Land-Cover Map Production. Implementation and optimization and validation of some algorithms in C++. Evaluation of their efficiency.

February - Mai 2011

Software Engineer at Accenture Technology Solutions

Development of a C# application to help after-sell service to follow up with incoming requests. Connection with an Access data base.

September 2008 - September 2010

Volunteer Missionary for the The Church Of Jesus Christ of Latter-Day Saints (Non-profit worldwide organization)

Human relationship, conflict management, planning and monitoring of objectives, teaching techniques training.

February - June 2008

Research Engineer at Centre National d'Études Spatiales (CNES)

Development of an high-resolution, satellite images segmenting application, in preparation for the Pleiades satellites. Study and implementation of SVM (Support vector machine) with the C++ library Orfeo ToolBox. Validation of results and documentation.