AMD Contributes ONNX Runtime Backend To FFmpeg DNN Filter
Phoronix
AMD has added an ONNX Runtime backend to FFmpeg's DNN filter, enabling hardware-accelerated AI inferencing via GPUs and NPUs.Summary
An AMD engineer, Steven Xiao, has contributed an ONNX Runtime backend to the FFmpeg Deep Neural Network (DNN) filter. This integration allows FFmpeg to utilize hardware acceleration for AI tasks such as upscaling and object detection. By leveraging the ONNX Runtime, the update enables support for various platforms, including NVIDIA CUDA, Windows DirectML, and notably, AMD's Ryzen AI NPUs via the VitisAI execution provider.
(Source：Phoronix)