Visit Xilinx at CVPR 2019 to learn about Adaptable Machine Learning Inference Acceleration.
The following showcases and demos will be featured throughout the show in the Xilinx booth # 860.
Demo #1 – ADAS Perception Acceleration
Multi-channel neural network processing for object detection with full automotive roadway scene segmentation and object detection and recognition. Demonstrates deep learning using CNN accelerators for a performance optimized ADAS solution. This demo uses the Zynq Ultrascale+ based ZCU102 platform running multi-neural network with four external cameras and loaded video stream.
Demo #2 – 8-ch Video Surveillance Appliances
Video analytics appliances are breathing new life into existing security systems by taking inputs from existing cameras and adding intelligence at the network. This demonstration shows how Xilinx enables simultaneous decode of eight H.264/H.265 video channels from different cameras and applies flexible and responsive AI in a single Zynq UltraScale+ MPSoC. Eight IP cameras (Xilinx or 3rd party) are sending compressed bitstream using a RTSP client to the appliance.
Demo #3 – High Density Video AI Applications
Aupera will be demoing real time camera feeds AI application for face detection, object classification as well as video transcoding and stream mixing on the same FPGA platform.
Demo #4 - Mixed Precision Inference on Alveo
Low-precision becomes popular in Deep Learning inference and FPGA benefits a lot. This demo deployed a flexible DPU IP on Xilinx Alveo Board with the support of mixed precision operation (INT8/4/2) and the mainstream of CNN models, which offers 2~3x performance compared to INT8. This demo shows the example of ResNet-50.
Xilinx is hosting a hands-on developer lab at CVPR.
Xilinx AI Inference on AWS F1
Wednesday, June 19 | 8:30 AM – 11:30 AM
Location: Westin Long Beach (located across the street from the Long Beach Convention Center)
In this lab session attendees will complete an end-to-end machine learning workflow, including network model, compilation, quantization, and deployment onto a Xilinx FPGA instances in AWS.