I-DEEP (Intempora Distributed Execution & Evaluation Platform) is a web-based application server for automating your algorithms test and validation procedures.
I-DEEP makes it easy to:
- store and retrieve resources (algorithms, models, simulation environments and scenarios, recorded sensors datasets...)
- validate your algorithms against a large range of pre-recorded sensors datasets or simulated scenarios
- test various configurations of your algorithms and find out the best options
I-DEEP can run distributed and automated tests putting together various software toolchains such as RTMaps of course, but also simulators, etc. Any software which is part of your algorithms validation toolchain can be integrated thanks to the development of a simple connector to I-DEEP.
ADAS Functions validation with I-DEEP
Advanced Driving Assistance functions will be expected to be more and more reliable until they can take full control on driverless cars. High-level functions such as image processing or decision-making to overtake some obstacle have to be tested extensively. Not only do they need to be bug free (where well known techniques can be used for validation such as formal code analysis, model checking, MISRA coding rules and alike, ISO 26262...), but it is also required that they have a wide range of operation and they can behave correctly against numerous complex scenarios. Automating the tests becomes compulsory and such tests have to be run with large amounts of pre-recorded sensors data but also with the help of advanced simulation tools.
Below is a simple example where the ProSIVIC simulator and the RTMaps middleware are used in the I-DEEP server to test the robustness of a dense stereovision algorithm from IFSTTAR/LIVIC against foggy weather conditions.
The simulator provides fog modeling on the stereovision cameras as well as the ground truth information in order to be able to easily and automatically benchmark the stereovision algorithm output. The fog density varies automatically on the successive test cases and the system can automatically store the amount of false detections, non-detections, and the detected distances errors.
This application has been demonstrated at the 2014 edition of the SafeyAssist conference in Aschaffenburg. The corresponding conference presentation can be downloaded clicking on this link.
The Pro-SIVIC simulator and an IFSTTAR dense stereovision algorithm integrated in RTMaps both operated automatically in the I-DEEP framework for validation of the algortihm against foggy weather conditions.