LogicTronix is “AMD-Xilinx Certified Partner“ & “Design Service Partner of Xilinx Kria SoM for AI/ML”.

We and Xilinx organized a Webinar on “AMD-Xilinx Kria-SoM Development and Machine Learning Acceleration” : AMD-Xilinx Kria SoM Webinar [Available On-Demand].

We offer following services to our Customers:

  1. Turnkey Solution development for Computer/Machine/Industrial/Medical Vision, ADAS, Security and Surveillance applications  – 8k/4k/2k based single/multi channel video processing capability with Xilinx FPGAs, MPSoC/SoC’s or Versal with Baremetal | FreeRTOS | Petalinux or OpenAMP flow.
  2. Turnkey Solution development for ML Acceleration with Xilinx Zynq SoC/MPSoC , Kria SoM  FPGA and Versal Adaptive SoCs:
    • AI/ML Model development (Dataset Preparation, Neural Network customization, Training, Pruning, Quantizing-Vitis AI, Compiling-Vitis AI) and MPSoC/SoM deployment.
    • We work with Caffe/Tensorflow/Darknet , Pytorch framework and
    • We use Vitis AI methodology for Quantization/Pruning and Compile the Model for specific MPSoC FPGA Board (Evaluation Board-EVK or Custom Board).
    • We develop custom solutions with Xilinx VVAS for Kria SoM-KV260 and MPSoC FPGA devices. We offer Multichannel/Multiple Model AI inferencing solution with Xilinx MPSoC FPGA and Kria SoM
    • Developing Solutions as Kria Apps for Kria KV260 and MPSoC boards [CG, EG or EV devices]. 
    • Know more on Machine Learning Acceleration –  “Machine Learning Portfolio – Slide Deck” or Visit: LogicTronix-ML-Solutions.
  3. Creating Multi channel Video Streaming System/Product with Xilinx MPSoC FPGA and Kria SoM. 
  4. FPGA Design and Development with RTL/HDL Flow | HLS flow |IP Flow | OpenCL Flow.
  5. IP Design & Verification and Design Migration – all families of Xilinx FPGAs
  6. Complete Linux based product development – GUI (GTK/QT for MPSoC/SoC) Development & Linux Driver Development 
  7. Crypto-Algorithm & Blockchain implementation on FPGA [High End FPGAs]
  8. Developing ultra-low latency designs or IP Cores for Finance (HFT), Medical and Mission Critical applications.
  9. With our PCB Partners, we also offer custom PCB or Carrier(for SoM) design services on MPSoC for ML/Computer Vision Solutions/Products. So we offer “idea into product” turnkey design services.
Embedded Linux based Development: Petalinux | Yocto | Buildroot
  1. Embedded System development with SoC FPGA’s.
  2. Linux driver development, Tcl Scripting & Bash Scripting
  3. Petalinux Development: Developing Petalinux Applications as Gstreamer-app for video streaming with ML acceleration
  4. Petalinux Development: developing solution with Multi threading, multi-processing , implementing Baremetal+Petalinux flow on OpenAMP. Developing time/latency critical applications and solutions.
  5. FreeRTOS Development for Computer Vision and Signal Processing Applications with Xilinx Zynq MPSoC and 7000 FPGA
Consultation and Design Migration:
  1. Consulting on FPGA Selection, Design Methodology and Product Development.
  2. Migrating Xilinx 6 series of FPGA design into Xilinx 7 Series FPGA.
  3. Inter-vendor FPGA design Migration: Migrating design from one to another vendor-tool or platform-  Xilinx, Intel & Microchip.
  4. Cost Optimized Platform selection and product development
Interfaces we mostly work :
  1. Computer Vision+ Machine Learning: MIPI, IAS, RTSP, USB , USB 3 Vision, SATA, GbE (Gigabit Ethernet), 10G ETH, SDI (3G-SDI, UHD, 12G-SDI), HDMI, TMDS/DVI, LVDS, Coexpress , SPI, I2C etc.
  2. ADAS-Automotive: GMSL2, FPD Link III, CAN, GbE (RADAR, LIDAR), SPI
  3. Medical Vision and Systems: GMSL2, SFP/SFP+, IAS, MIPI etc. 
Protocols we mostly work:
  1. AXI, Aurora, PCIe, Native Video, BT656, TMDS, SDI(SDI/3G-SDI/6G-SDI/12G-SDI), SFP, HDMI, DP, SPI, I2C, MIPI CSI/DSI etc.
FPGA Families and Devices we work with:
  1. We work with all families and devices of Xilinx (AMD-Xilinx)- ranging from cost optimized Artix/Spartan to Embedded hero – MPSoC (Ultrascale+) , high performance – Kintex/Virtex  and Accelerator’s – Alveo/Versal.
  2. Artix/Spartan 7 series to Ultrascale+ : we had worked for “Video processing and sensor processing products”  while Kintex/Virtex are targeted for high end  and high throughput applications. If you need to migrate Spartan 6 or older FPGA design to new one then you can contact us, we have good experience and expertise in it (to migrate cost optimized way).
  3. Zynq Ultrascale+ MPSoC: it is most suitable for Vision, Edge and AI applications. We have developed and worked on multiple solution with this family.
  4. Alveo: it is cloud centric (datacenter) FPGA platform. We have worked with AWS F1, Nimbix (now not available), VMAccel based platform for different workload acceleration.
  5. Versal : it is fulfilling customer requirement which is not easy to fit in MPSoC. Versal is one of recently lunched variant of FPGA family, we already have worked with AMD-Xilinx for some development (some will release in 2025) with it for some of customer’s requirement.

We are also experienced with Microchip and Intel. For those vendor based development or inquiries, please write us at email – info@logictronix.com.

Machine Learning Acceleration flow and Framework:

We offer turnkey solution on Machine Learning Acceleration including,

Dataset Preparation --> Training--> Network Optimization--> Pruning--> Quantization--> Compilation--> Xmodel deployment on MPSoC Boards.

We also offer custom PCB or Carrier(for SoM) design services on MPSoC for ML/Computer Vision Solutions/Products.

We work with all version of AMD-Xilinx Vitis AI tools (also have good experience with DNNDK which is now legacy though). Here are the some of popular ML framework we work for ML Acceleration on MPSoC/Versal.

  1. Caffe/Darknet : Large number of Vitis AI Models at Model Zoo are in Caffe and we do have sound expertise with caffe based training, optimizing and customizing neural network and converging model with Vitis AI. Due to vast number of models, caffe is highly preferred framework with Vitis AI, even though there is very less resources on caffe publicly available.
  2. Tensorflow (TF1/TF2): We work with tensorflow framework for training, quantization , pruning and compilation to model deployment on MPSoC Board.
  3. Pytorch: Pytorch is one of popular ML framework , we work with pytorch for ML acceleration on MPSoC and Versal Platforms.
  4. ONNX/TVM based ML model development and acceleration: ONNX is popular for interoperability of ML models, we also use ONNX for deployment of ML models in edge FPGA Boards (MPSoC).
AI/ML Acceleration flow on FPGA

AI/ML Acceleration flow on FPGA – LogicTronix

 
Machine Learning Acceleration with Versal

LogicTronix also work with AMD Versal Platforms for ML acceleration on Edge, if you like to know more in it then please check:
https://logictronix.com/machine-learning-with-fpga/ml-acceleration-with-amd-versal/

Contact Us:

For design services and products related inquiry/information, please fill-up the following form. We will contact you from info@logictronix.com!