
Raven: A 100% GPU-driven AI inference framework for real-time video and graphics
Learn how Raven redefines real-time inference with a fully GPU-driven architecture for multimedia and video pipelines.
Discover how Fluendo has shaped the digital landscape through our dedication to innovation and quality in multimedia.
We specialize in providing multimedia products for multiple frameworks and platforms, including audio and video codecs, multimedia SDKs, image processing modules, video streaming plugins, AI components, and more.
We work with GStreamer to deliver exceptional multimedia experiences and our years of expertise guarantee outstanding consulting services for different industries and business sizes.
Our extensive range of plugins and solutions ensures seamless playback of any audio or video format across virtually any sector and industry.
Our experts handle all potential challenges, saving you from worries about development time, project costs, or technical difficulties.
With 20 years of experience, Fluendo has been delivering top-tier services to small to the biggest and finest tech companies worldwide.
We specialize in key industries like Broadcasting, DaaS & VDI, Sports, and Video Surveillance, delivering high-performance media experiences. Our expertise spans GStreamer, application development, codecs, browsers, Multimedia Edge AI, and more.
Fluendo enhances media performance with cutting-edge products: the Fluendo Codec Pack for seamless playback, the FFmpeg Enabler for extended compatibility, and Fluendo AI Plugins for intelligent multimedia processing.
Fluendo offers consulting services for advanced multimedia solutions, including GStreamer optimization, custom training, and CPU usage reduction. We help streamline audio and video workflows, ensuring the success of your multimedia projects.

Learn how Raven redefines real-time inference with a fully GPU-driven architecture for multimedia and video pipelines.

Table of contents NMS-Raster: Post-processing bounding boxes using the “G” in GPU The bounding box problem Classical NMS NMS-Raster: Z-Aware rasterized suppression Advantages of the approach Implementation Instanced drawing into integer framebuffer Contribution histogram Threshold filtering Strengths of the design Performance NMS-Raster: Post-processing bounding boxes using the “G” in GPU In this article, we explore the challenges of bounding box post-processing in AI-powered object detection and demonstrate how our high-performance inference engine, Raven, addresses them from a novel angle.

I promise I won’t use ML to write the blog this time. Just my Emacs and me, and a little copy-paste. The GStreamer Conference 2025 is just around the corner, and Fluendo is excited to be a proud sponsor and presenter once again. This year, the event will take place from October 23 to 26 in London, UK, bringing together industry experts, developers, and enthusiasts to share the latest advancements in multimedia frameworks.

Discover Fluendo's outcomes from IBC 2025 in Amsterdam.