Multi-task models play a pivotal role in autonomous vehicle development by enabling developers to perform multiple critical functions simultaneously. These models, with multi-task heads generating application-specific outputs, require careful deployment to maximize compute resource efficiency.

In this webinar, we demonstrate how:

• Multi-task models can be deployed on NVIDIA DRIVE Orin by leveraging different compute devices on the SoC at the same time, to improve latency and throughput
• To implement a sample inference application using the NVIDIA software stack, including CUDA, TensoRT, and cuDLA

Content

content goes here