This release is a collaborative effort with Open Robotics to bring new features of the Humble release to boost hardware-acceleration performance and improve perception tasks.
NVIDIA’s implementation of type adaptation (REP-2007) and type negotiation (REP-2009) is called NITROS (NVIDIA Isaac Transport for ROS) which reduces system software/CPU overhead and helps gain higher throughput.
We also introduce two new DNN-based GEMs:
- ESS to perform stereo depth estimation via continuous disparity prediction.
- Bi3D for vision-based obstacle prediction.
Additional resources:
Blog: Improve Perception Performance for ROS 2 Applications with NVIDIA Isaac Transport for ROS
Youtube: Isaac Stereo Perception
Youtube: Type Adaptation Video