An AMD engineer has contributed to the upstream FFmpeg library an ONNX Runtime back-end for its DNN filter. The FFmpeg Deep Neural Network (DNN) filters allow for running AI models natively inside the video processing pipeline for upscaling, object detection, background segmentation, and more. This ONNX Runntime back-end support is notable in that it expands the GPU and NPU capabilities with FFmpeg...
The increasing prevalence of AI models and the demand for efficient, integrated AI processing within existing workflows (like video) necessitates closer hardware-software integration.
This development streamlines the deployment and inferencing of AI models directly within a widely used media processing framework, making AI capabilities more accessible and performant for a broad range of applications.
FFmpeg gains enhanced capability to run diverse AI models with better hardware acceleration across different vendors, potentially lowering the barrier for AI integration in media workflows.
- · AMD
- · Video Processing Software Developers
- · AI Model Developers
- · Media & Entertainment Industry
- · Proprietary AI video processing solutions
- · CPU-only AI inference workflows
FFmpeg users can now natively integrate and accelerate a wider array of AI models for tasks like upscaling and object detection within their video processing pipelines.
This improved integration could lead to the proliferation of AI-enhanced video content and tools, democratizing advanced AI capabilities for content creation and analysis.
The enhanced capability for on-device AI video processing could reduce reliance on cloud-based solutions for certain tasks, impacting data transfer and cloud computing demand.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at Phoronix