How to harvest reliable AI training data for sensor function validation
Harvesting real world data in test drives is crucial moving forward validating sensorics functionality. Developing safe automated driving systems, we can trust, requires training data that is reliable, consistent and accurate. At the same time sensor resolutions and thus also the bandwidth increases in every project.
The challenge of bandwidths can be solved with a flexible data harvesting concept based on data center technologies.
The accuracy and consistency of harvested data is achieved by data integrity monitoring and time correlation of sensor data streams. So, covering camera, radar and Lidar raw data enriched with bus and meta data, a holistic view of a test drive can be acquired in accurate quality with no bit lost.
3 MAIN TAKEAWAYS
– Engineers get an insight, what challenges data harvesting on the road occur and what process steps are important
– Perception Engineers get an insight on how a vehicle measurement setup can be expanded during a project
– Developers of AI Based systems see how Data Quality is already influenced during data acquisition
An AutoSensPLUS subscription is required to watch this video on demand.
Presentations, panel discussions, interviews and more from all AutoSens events, with brand new content streamed both live and on demand.
Want to watch this FREE session?
You must be logged in to watch this session.
Please either login or create an account below.
Already a subscriber? Please Log in
More on the topic
More from the partners