DAISY - An Eclipse based IDE for Automotive Machine Learning | EclipseCon Europe 2018
4EclipseCon Europe 2018
Ludwigsburg, Germany · October 23 - 25, 2018
https://www.eclipsecon.org/europe2018
Speaker(s):
Andreas Graf (itemis AG)
Judith Schaeffer (Continental Automotive GmbH)
https://www.eclipsecon.org/europe2018/sessions/daisy-eclipse-based-ide-automotive-machine-learning
Teaser: Today, the development of automotive software functions based on artificial intelligence still needs a lot of different tools, languages and frameworks.
In this talk, we introduce how we built upon the Eclipse eco-system an integrated development environment called "DAISY" (Development environment for Artificial Intelligence SYstems).
Developers of machine-learning (ML) based functionality use a lot of different data and SW sources, languages (Python, C++), Frameworks (Tensorflow, Theano, Numpy) and development tools (cmake, anaconda) and infrastructure (Docker, AWS). But there is still a lack of a seamless tool support by an integrated development environment, leaving developers with manual tasks for all the heterogeneous steps. In our talk, we will introduce the use cases that we cover with DAISY, such as:
* Integration of data from different sources, either locally, remote or cloud.
* Neural Network development, testing and visualization with different frameworks, both locally and in the cloud (for example object detection, product inspection, etc.)
* Integrating neural networks on a prototyping system (such as Nvidia Drive PX) by rewriting in C, cross-compilation and dockerization.
* Managing different configurations of development systems locally.
We will then introduce how we support these use cases by using projects from the Eclipse eco-system and extending it with custom plugins, e.g.
* Python Development
* Conda Environment Management
* C++ Development
* Remote Execution and testing (in the Cloud, on AWS)
* Dockerization
* Execution on target platform (Nvidia Drive PX)
At the end of the talk, the audience will have learnt about how functionality for autonomous cars is being developed and how Eclipse can easily be leveraged to support the development.