Objective:
Designed and Developed for one of the leading Live Streaming solutions provider of Europe.
The project objective was to create an SDK / Library that could be used by the application to simultaneously decode multiple channels of HEVC compressed streams at real-time.
The application is a vision mixer software which is required to display a preview of the encoded content from multiple live camera feeds in a sports arena.
Client: A leading Live Streaming solutions and Live Production tools provider of Europe.
Technical Challenges:
- Real-time decoding of multiple channels of HEVC streams with minimum latency and no delay amongst the multiple camera feeds was a big challenge.
- Minimum CPU Utilisation: The Decoding engine for the channels was required to consume only minimal CPU resources, leaving enough compute power and memory for the clients’ main streaming application.
- Efficient Video rendering: The video rendering was required to be efficient and following both open and proprietary standards.
Solution:
- Real-time decoding of 8 Full-HD channels was achieved by using NVIDIA CUDA and Hardware decode technologies.
- With the GPU taking away the major decode engine task load, the CPU was virtually left free for the rest of the application tasks.
- OpenGL and Direct3D based rendering techniques were employed to eliminate the CPU – GPU data transfer overheads.
Technologies / Tools used:
- NVIDIA cards supporting NVDEC (Hardware Video decoding)
- NVIDIA CUDA Technology
- Direct 3D / OpenGL Video rendering engine
- Microsoft Visual Studio IDE
- Intel i7 / XEON / Windows / Linux
Outcome:
Customer was able to integrate the Decoder SDK / library with their vision mixer application. The decode performance of the engine perfectly matched their latency requirements. Also, the integration of HEVC decode capabilities enhanced the end product's value by adding features to playback / preview HEVC encoded content.
Do write to us at sales@intecoretech.com to know more about our NVIDIA GPU based 8-Channel HEVC Decoder case study.