Customer: A US-based leading start-up company developing technology for self driving car
Scope:
The customer was looking for a partner to support in autonomous vehicular validation in a lab-based environment.
Solution:
- Platform: Linux, Nvidia Quadro GPUs
- Implemented following inputs to the system:
- H.265 encoded video file
- Mapping from frame index to timestamp for each H.265 file
- Clock signal indicating the current time
- A mapping from video file to display number
- Enabled playback of a set of logged camera data (for multiple cameras) in sync with each other and other time stamped data (which will be played back separately)
- Enabled video files to be played on correct output according to the configuration file along with additional meta data to be encoded in each frame
- Decoding logs containing images from multiple cameras from H.265 and rendering to separate display port outputs across multiple GPUs
Impact:
Sasken leveraged its expertise in sensor fusion, platform engineering and multimedia to achieve zero frame drop with synchronized playback.
