GstCUDA – CUDA Acceleration Framework for GStreamer Multimedia Pipelines
GstCUDA by RidgeRun is a robust GStreamer-based framework that simplifies the integration of custom NVIDIA CUDA algorithms into multimedia pipelines. Designed for embedded platforms like NVIDIA Jetson, GstCUDA allows engineers to focus exclusively on their CUDA code—without worrying about how GStreamer manages buffers, synchronization, or plugin architecture.
By abstracting away the boilerplate required to build GStreamer elements, GstCUDA accelerates development, enhances performance, and streamlines deployment for GPU-accelerated video and image processing.
Key Features:
-
Focus on CUDA, Not GStreamer
Developers write only the CUDA algorithm logic—GstCUDA handles memory management, pipeline integration, and GPU buffer flow. -
Base Classes for Plugin Development
Includes SIMO, MISO, MIMO, and filter templates to easily create new GStreamer plugins for your CUDA operations. -
Zero-Copy NVMM Support
On Jetson platforms, supports zero-copy GPU memory to eliminate CPU-GPU transfer overhead. -
Runtime CUDA Kernel Loading
Dynamically load custom CUDA libraries during runtime for rapid experimentation and modularity. -
Prebuilt Plugins Available
Includes ready-to-use GPU elements likecudadebayer,cudaawb,cudashift, and more—built on top of GstCUDA.
Example Use Case:
An industrial inspection system using NVIDIA Jetson Xavier NX needs to apply real-time defect detection and noise reduction. With GstCUDA, the engineering team focuses purely on implementing the CUDA-based image filters. GstCUDA handles all buffer movement, memory mapping, and GStreamer plugin setup, allowing seamless integration into the existing video pipeline—achieving high performance with significantly reduced development effort.
What’s Included:
-
GstCUDA plugin framework
-
Base classes for rapid plugin development
-
Sample elements and usage examples
-
Full documentation and API references
-
Optional 11-hour Embedded Software Support package
Supported Platforms:
-
NVIDIA Jetson Nano, TX2, Xavier NX, AGX Orin
-
x86-64 Linux with NVIDIA GPUs
-
Any embedded Linux system with CUDA + GStreamer 1.x
Why RidgeRun?
With over 19 years of experience in Embedded Software Consulting, RidgeRun builds efficient and scalable frameworks for multimedia, AI, and vision pipelines. GstCUDA helps you get the most out of NVIDIA’s hardware without re-implementing GStreamer integration for every new algorithm.
Learn more in the developer wiki
Explore the GstCUDA Framework
See real pipeline examples