Gst-mfx

0 %
Loading Click here to add:
Add to notification list

My favorites With attachments Unanswered New My annotations
Reset filters
Get notified of changes by email

This talk presents GST-MFX, a set of GStreamer plugins that uses Intel Media SDK (MSDK) to perform high performance decode, encode, video post-processing (VPP) and rendering for both Windows and Linux platforms.

 

GST-MFX addresses a key segment of application developers who wish to develop media-based cross-platform applications for both Windows and Linux while getting the very best performance from utilizing MSDK, which in itself is a sophisticated, proprietary media API that is highly-optimized for Intel's platforms starting from 3rd generation Intel Core processors. MSDK is also very well-supported and has comprehensive documentation on its capabilities all the way down to low-level codec performance finetuning.

 

We first intend to talk about the motivation of this work, and then the very promising value it brings to the table to the GStreamer ecosystem. We then discuss some of best features that GST-MFX has, and compare it to existing alternative GST-MSDK implementations, as well as GST-VAAPI.

 

Finally, we will briefly summarize the architecture and implementation details which allowed us to achieve the current state of GST-MFX, and where it stands in terms of production-level software. You can check out the latest ongoing development of GST-MFX from here:

 

https://github.com/01org/gstreamer-media-SDK/tree/topic_linux_and_window

https://github.com/ishmael1985/gstreamer-media-SDK

https://github.com/ph0b/gstreamer-media-SDK

Ishmael Sameen is a former Intel software developer and currently a PhD research student at University Paris XIII, whose main interests lie in doing cool stuff and applying real-world solutions in the area of image and video processing. He has worked (and is still working with a few) with various technologies such as nVidia CUDA, Altera FPGAs, FFMPEG, QT, Intel Media SDK, etc., but only recently touched GStreamer two years ago for a customer requirement (after trying to unsuccessfully convince them to use FFMPEG). Since then, he enjoys working with the GStreamer framework and learning from their codebase, even after having left Intel to pursue his research interests in neural networks for computer vision.

Xavier Hallade is an Application Engineer at Intel, focused on improving 3rd party applications performance and features for upcoming PC platforms. He’s been helping customers integrate video hardware acceleration in their Windows applications several times over the past year, and decided to scale this work by contributing to the GStreamer framework.