The vvas_xinfer plugin is designed to perform hardware-accelerated inference on video and images using various neural network models.
This plugin is based on the ONNX Runtime and VART inference engine. ONNX Runtime supports both CPU Execution Provider (CPU-EP) and the VitisAI Execution Provider (VitisAI-EP), as well as mix of CPU-EP and VitisAI-EP. Vitis-AI-Runtime (VART) provided alternatives to bypass inefficiencies in acceleration further to extract HW maximum acceleration.
It integrates seamlessly with GStreamer pipelines, enabling efficient use of hardware accelerators or CPU for real-time video analytics.
Whether you are working with object detection, classification, segmentation, or custom neural networks, vvas_xinfer provides a flexible and powerful solution.