The Vitis™ AI platform is a comprehensive AI inference development solution for AMD devices, boards, and Alveo™ data center acceleration cards. It consists of a rich set of AI models, optimized deep learning processor unit (DPU) cores, tools, libraries, and example designs for AI at the edge and in the data center. It is designed with high efficiency and ease of use in mind, unleashing the full potential of AI acceleration on Xilinx FPGAs and adaptive SoCs.
Figure 1 - Vitis AI Structure
Figure 2 - Model Zoo
AI model zoo is open to all users with rich and off-the-shelf deep learning models from the most popular frameworks such as Pytorch, Tensorflow, Tensorflow 2, and Caffe. AI Model Zoo provides optimized and retrainable AI models that enable faster execution, performance acceleration, and production on AMD platforms.
With exceptional model compression technology, the AI optimizer reduces model complexity by 5X to 50X with minimal accuracy impact. Deep compression takes the performance of your AI inference to the next level.
Figure 3 - Vitis AI Optimizer
Figure 4 - Vitis AI Quantizer
A completed process of custom operator inspection, quantization, calibration, fine-tuning, and converting floating-point models into fixed-point models that requires less memory bandwidth - providing faster speed and higher computing efficiency.
The AI compiler maps the AI model to a highly efficient instruction set and data flow. It also performs sophisticated optimizations, such as layer fusion and instruction scheduling, and reuses on-chip memory as much as possible.
Figure 5 - Vitis AI Compiler
Figure 6 - Vitis AI Profiler
The performance profiler allows programmers to perform in-depth analysis of the efficiency and utilization of the AI inference implementation.
The Vitis AI Library is a set of high-level libraries and APIs built for efficient AI inference with DPU cores. It is built based on the Vitis AI Runtime (VART) with unified APIs and provides easy-to-use interfaces for AI model deployment on AMD platforms.
Figure 7 - Vitis AI Library
Figure 8 - Vitis AI Compiler
The WeGO framework inference flow offers a straightforward path from training to inference by leveraging native TensorFlow or PyTorch frameworks to deploy DPU unsupported operators to the CPU—greatly speeding up model deployment and evaluation over cloud DPUs.
Deep Learning Processor Unit (DPU)
The DPU is an adaptable domain-specific architecture (DSA) that matches the fast-evolving AI algorithms of CNNs, RNNs, and NLPs with the industry-leading performance found on Zynq™ SoCs, Zynq UltraScale+™ MPSoCs, Alveo data center accelerator cards, and Versal ACAPs.
Figure 9 - Vitis AI DPU
The Vitis™ AI platform delivers powerful computing performance with the optimal algorithms for edge devices while allowing for flexibility in deployment with optimal power consumption. It brings higher computing performance for popular edge applications for automotive, industrial, medical, video analysis, and more.
Empowered by the Vitis AI solution, Alveo™ data center accelerator cards offer competitive AI inference performance for different workloads on CNNs, RNNs, and NLPs. The out-of-the-box, on-premise AI solutions are designed to meet the needs of the ultra low-latency, higher throughput, and high flexibility requirements found in modern data centers—providing higher computing capabilities over CPUs, GPUs and lower TCO.
Working with public cloud service providers such as AWS and VMAccel, AMD now offers remote access to FPGA and Versal™ ACAP cloud instances for quickly getting started on model deployments—even without local hardware or software.
Extensive documentation support is available for developing with the Vitis™ AI platform on models, tools, deep learning processor units, etc.
Link to specific documents below, or visit the Documentation Portal to see all the Vitis AI Platform documents.
With Vitis™ AI, it is now possible to achieve real-time processing with the 3D perception AI algorithm on embedded platforms. The co-optimization from hardware and software speed up delivers leading performance of the state-of-art PointPillars model on Zynq® UltraScale+™ MPSoC.
Latency determines the decision-making for autonomous driving cars when running at high speeds and encountering obstacles. With an innovated domain-specific accelerator and software optimization, Vitis AI empowers autonomous driving vehicles to process deep learning algorithms with ultra-low latency and higher performance.
With strong scalability and adaptability to fit across many low-end to high-end ADAS products, Vitis AI delivers industry-leading performance supporting popular AI algorithms for object detection, lane detection and segmentation in the front ADAS, and In-cabin or surround-view systems.
Cities are increasingly employing intelligence-based systems at the edge point and cloud end. The massive data generated every day requires a powerful end-to-end AI analytics system in order to quickly detect and process objects, traffic, and face behavior. This adds valuable insight to each frame from edge to cloud.
Learn more about Xilinx in Machine & Computer Vision >
Develop accelerated applications with the Vitis AI platform in the cloud—no local software installation or upfront purchase of hardware platforms necessary (pay as you go). Log in and get started right away.