NVIDIA DeepStream SDK

Removing the Boundaries of Video Analytics.


DeepStream SDK delivers a complete toolkit for real-time situational awareness through intelligent video analytics (IVA). With hardware-accelerated building blocks, the application framework allows you to focus on building core deep learning networks and IP rather than designing end-to-end solutions from scratch.

The SDK supports a diversity of use cases, using AI to perceive pixels and analyze metadata with the flexibility to span from the edge to the cloud. This includes retail analytics, intelligent traffic control, automated optical inspection, freight and goods tracking, web content filtering, ad injection, and more.

Rocket Fuel

For complex perception problems, developers can use the DeepStream SDK’s heterogeneous concurrent neural network architecture to better understand video content. Different deep learning techniques deliver more intelligent insights and detailed attributes — like an object’s color, brand, or association.

DeepStream makes use of:

  • NVIDIA® TensorRT and NVIDIA CUDA® for AI and other GPU computing tasks
  • Video CODEC SDK and multimedia APIs for accelerated encode and decode
  • Imaging APIs for capture and processing
  • A graph-based architecture and modular plugins to create configurable processing pipelines

What's New in DeepStream SDK 3.0

The DeepStream SDK 3.0 for NVIDIA Tesla® introduces new capabilities that make it much simpler to build and deploy complex AI processing pipelines.

  1. A modular plugin approach helps develop highly flexible apps, while containers offer straightforward deployment and upgrades.
  2. New plugins connect the edge to the cloud through message brokers for analytics with custom deployments based on Spark or by using IoT services.
  3. Pruned and efficient model support created by NVIDIA’s Transfer Learning Toolkit helps tightly pack complex applications that deliver high throughput and stream density.

The SDK also introduces support for the NVIDIA Turing architecture, dynamic stream management, multi-GPU processing, dewarping capabilities for processing 360-degree cameras, and a number of plugin enhancements.

DeepStream SDK 3.0 plugins

The new SDK includes the following hardware-accelerated plugins:

  • H.264 and H.265 video decoding
  • Stream aggregation and batching
  • TensorRT-based inferencing for detection and classification
  • Object tracking reference implementation
  • On-screen display API for highlighting objects and text overlay
  • Frame rendering from multi-source into a 2D grid array
  • Accelerated X11/EGL-based rendering
  • Scaling, format conversion, and rotation
  • Dewarping for 360-degree camera input
  • Metadata generation and encoding
  • Messaging to cloud
  • On-screen display API for highlighting objects and text overlay
  • Frame rendering from multi-source into a 2D grid array
  • Accelerated X11/EGL-based rendering
  • Scaling, format conversion, and rotation
  • Dewarping for 360-degree camera input
  • Metadata generation and encoding
  • Messaging to cloud

DeepStream in Docker

Applications built with the DeepStream SDK can now be deployed using a Docker container, enabling flexible system architectures and straightforward upgrades that greatly improve system manageability. The Docker containers will be available on NVIDIA GPU Cloud (NGC).

DeepStream also comes with sample code and pre-trained deep learning models as examples to perform classification and object detection on video streams.

DeepStream is available for NVIDIA Tesla and Jetson. Download the DeepStream SDK today to get started.


NVIDIA Transfer Learning Toolkit

Transfer Learning Toolkit, along with DeepStream SDK 3.0, offers an end-to-end deep learning solution for Intelligent Video Analytics on Tesla GPUs. You can now accelerate development of efficient deep learning networks and prune networks to tightly pack complex applications, delivering high throughput and stream density.


Blogs

Webinar