Training and executing neural networks with large amounts of data requires enormous computing power, which is typically found in a data center. Frameworks and software like Tensorflow, Caffe or Matlab, libraries like Pandas, Numpy or Pytorch paired with container technologies like Docker are set up and work in the lab. So far everything is clear!
But in the vehicle?
There you will face new challenges of a different kind. Suddenly, limited installation space, extreme ambient temperatures, as well as strong vibrations and shocks play a role that server systems cannot meet.
DATALynx vehicle servers
Our DATALynx series is designed to meet the difficult balancing act between performance and robustness during testing and validation of sensor systems. It is specifically designed for demanding in-vehicle applications, while delivering the exceptional performance expected in the data center. Whether it's a megabrain for AI and image processing or a data kraken for recording with many interfaces.
We bring worlds together with the DATALynx Series and unleash the power onto the road.
Prototyping algorithms and processing images is often needed during the first steps of developing a sensor based system. An additional task would be live pre-labeling and filtering of scenarios live during test drive. See following simplified setup for online analysis and processing.
The all-rounder: The combination of concentrated performance with integrated memory for data logging
The data aggregate: Acquisition and storage of large amounts of data at data center speed
The intelligent one: Configuration for up to four dedicated graphic cards for parallel computing, image processing and neuronal networks.
More information about our AVETO Toolchain
Replay | Simulation of sensor raw data