, Product Manager - Metropolis AI Workflows, NVIDIA
, Sr. CUDA Math Library Engineer and Team Lead, NVIDIA
, Lead Technical Product Manager, NVIDIA
The field of Vision-AI demands tools for building scalable, high-performance applications that harness state-of-the-art GPU-powered inference. There are two primary approaches to develop such applications: - Utilizing APIs for hardware-accelerated libraries - Employing software frameworks.
This presentation will explore the trade-offs between these two approaches and introduce NVIDIA DeepStream Libraries a new feature that adds supports both methods for the first time. Discover how the new DeepStream Libraries Python APIs supercharge developers seeking to tap into GPU-accelerated Vision-AI capabilities without a framework dependency.
You will also learn about the latest new DeepStream SDK's that leverage GStreamer framework: a novel programming model that simplifies GStreamer usage, REST-APIs for pipeline management, support for sensor fusion with the BEVFusion model, and a developer tool for optimizing DeepStream pipelines. Join us to advance your Vision-AI development with the latest DeepStream innovations.