We are going to have different Case Studies on “Vision Systems, Automotive Design and Machine Learning Acceleration” of LogicTronix at this Page!
A. Case Studies- Vision Systems
1. High Throughput Video Processing -4K on ZCU102 MPSoC Board
LogicTronix has created the “Video System Pipeline” in VIVADO targeting 4K Video Processing and Mixing on Standalone/Baremetal and Petalinux Systems. This video system implemented on Xilinx ZCU102 MPSoC Board for prototyping take 4K@30 resolution of HDMI stream, combine it with two different 1080p@30 resolution of HDMI stream came from two FMC interface and generate the processed output at 4K@30. This system is also applied for Medical Vision System design and development for our customer and also used for other industrial customers for high resolution video mixing application.
This design uses two FMC interface of Xilinx ZCU102 Board for “FMC-HDMI” from Avnet and Digilent. These FMC HDMI are used for taking HDMI input and sending HDMI output respectively. Following picture shows the test setup plan for “high throughput video system on ZCU102”.
2. Xilinx MIPI CSI2 Tx IP Test and Platform on Kria KV260
Most of the vision system design use “MIPI CSI2 Rx (for MIPI Camera)” IP Pipeline for interfacing MIPI Camera and performing the ISP, other processing and send to Display Monitor. But there is also need of MIPI CSI2 Tx IP Core when we want to emulate Camera on FPGA device or when we want to send data via MIPI interface to another FPGA or processing platform.
Our MIPI CSI2 Tx test plan uses “Xilinx MIPI CSI2 Tx IP Core” and its Standalone driver on one MPSoC Board , we use MIPI ribbon cable to send data and receive it on another MPSoC/SoC FPGA Board.
We have develop and test this Tx-Rx design in Kria KV260 to KV260 and KV260 to Zybo , two different setup. In MPSoC to MPSoC or Kria-Kria we can harness MPSoC or Ultrascale+ FPGA architecture and its MIPI IP along with rich set of PL resources, high speed interfaces, DPU for Machine Learning and VCU for encoding and decoding, so it allows to develop different solution or application for vision systems. While in MPSoC/Kria to Zybo we showcased the flexibility of design to received at Zynq family of FPGA.
Following test setup shows the Xilinx Kria KV260 sending MIPI data/stream to Zybo Z7-10 SoC FPGA Board of Xilinx. Here we are implementing MIPI CSI2 Tx on Xilinx Kria KV260 and MIPI CSI2 Rx at Zybo Z710 Board. The MIPI interface of Kria KV260 is made or “Camera Interface(MIPI CSI2 Rx)” but we have twekded/hacked its MIPI Interface from design and physical connection to send the MIPI CSI2 Tx data.
3. Multi-MIPI Camera Design for Industrial and Automotive Application
LogicTronix FPGA Design team also have developed “multi-camera” based video processing pipeline in MPSoC (ZCU102) FPGA which did the MIPI data capture, all the ISP (Bayer2RGB, Gamma, Sharpness/Edge enhancement, AWB etc) , perform the feature matching and stitching application. This design is used for surround view system for Automotive and multi-camera image stitching for mapping application. In this system we have used Opsero’s (thanks to Jeff Johnson) RPi FMC which can take 4 MIPI Camera input and send to FPGA fabric via FMC. After capturing the data and ISP we also do the image encoding with our own custom JPEG Encoder IP. We have more than 7 IP cores for the Image Signal Processing(ISP) which allows to lower resource usage, design flexibility and enhancement on the output image data.
We also have similar multi-camera system designed with GMSL2 interface (GMSL2 camera and FMC) for Automotive Application.
Our IP Cores are also available at Xilinx IP Catalog: https://www.xilinx.com/alliance/memberlocator/1-1dturdk.html#products
Kudos to our Senior Engineer Nikil Thapa , FPGA Engineer -II Diwakar K.C and to our FPGA Design Team for Creating these Case Studies on “Vision System Development with Xilinx MPSoC/SoC and FPGAs”.
B. Case Studies- Machine Learning Acceleration
Case Studies Coming Soon! Please check our “White Paper, AI/ML-FPGA Page and Tutorials” or Please check our Hackster tutorials on ML Acceleration at: https://www.hackster.io/LogicTronix/
For any inquiry on our Case Studies or any for inquiry on your design service please write us an email to: info@logictronix.com!