Vitis™ AI is a comprehensive acceleration platform for machine learning inference development on AMD Xilinx platforms. It consists of a series of optimized IP, software tools, libraries, deep learning models from multiple industry-standard frameworks and sample designs, with which developers can easily achieve accelerated AI inference abilities on both FPGA and ACAP.

Supports mainstream machine learning frameworks, Pytorch, TensorFlow1 and TensorFlow2, and popular neural networks capable of diverse applications; Provides powerful AI Quantizer, Optimizer and Compiler that realize significant processing efficiency optimization; High-level C++/Python APIs for fast implementation on edge and cloud; Highly scalable DPU IP cores on Zynq SoCs, ACAPs and Alveo acceleration cards to realize lower latency, higher throughput or lower power consumption.