Vitis AI-ML on Sundance VCS3 with Prophesee Event Camera

The VCS3 is one of the smallest form-factor MPSoC platforms available on the market. Its compact size and low power consumption make it an ideal platform for edge AI applications. Powered by the XCZU3EG MPSoC, the VCS3 delivers a low-latency, power-efficient processing pipeline suitable for real-time AI workloads.

This article demonstrates how to create an FPGA design for the Sundance VCS3 MPSoC development kit, integrating the Prophesee IMX636 event-based vision sensor (aka, Neuromorphic or DVS sensor) and deploying a machine learning model to process event-based data in real time.

Prophesee event based vision sensor are neuromorphic sensor or dynamic vision sensor (DVS).

We are using 2022.2 tool version, VIVADO and Petalinux for this application development.

Quick Overview of this Application

VIVADO Pipeline - overview

The VIVADO pipeline consists of IMX636 MIPI Pipeline for Prophesee event sensor data acquisition, DPU hierarchy for ML inference and Display Pipeline for showing results of EventML at VCS3 3D kit.

We are using Vitis AI 3.0 and DPU B512 variant for ML inference. With these IP blocks in the design, following is the resource usage by the complete pipelines.

Petalinux Development

We developed the petalinux project and build artifacts for the VCS3 kit based on the template provided by Sundance. VCS3 has three major approach of booting Linux, i.e. JTAG, QSPI and eMMC.

We are booting BOOT.BIN, boot.scr and image.ub at one partition of eMMC and Rootfs at 2nd partition.

To acquire the Prophesee IMX636 event based sensor data, we have to use the metavision SDK (openEB core) on the Petalinux image. And the camera access, some of data handling is happening at Metavision SDK.

The same app including metavision SDK runs the DPU based ML inference and displaying the result on the display-monitor which came along with VCS3 3D kit.

Quick demo:

The machine learning model demonstrated above processes event-based data captured by the sensor in real time. With fewer than 400k parameters, the model is highly optimized for edge AI inference, enabling low-latency and power-efficient execution on the compact Sundance VCS3 MPSoC platform. This demonstrates that even a small form-factor MPSoC can effectively run real-time event-based vision applications without requiring high-end computing hardware.

Kudos to the LogicTronix FPGA, Embedded, and Machine Learning team for their dedication and expertise in developing this application. Special thanks to the Sundance team for their continuous support and collaboration throughout the project.

If you are interested in evaluating this application, please contact Sundance or LogicTronix. For inquiries, reach out to info@logictronix.com to discuss evaluation options, technical details, or collaboration opportunities.

Like to know more?
Check this Event based ML Application from AMD-Xilinx Appstore - https://www.amd.com/en/developer/resources/kria-apps/object-detection-and-tracking.html

LogicTronix is AMD-Xilinx Partner for FPGA Design and AI/ML Acceleration.

LogicTronix will be demonstrating its latest sensor fusion and Edge AI solutions built on AMD-Xilinx FPGA and Adaptive SoC platforms at the AMD Embedded+ Conference Series (ECS) events in Tokyo and Bangalore in August 2026. For more information about the event and demonstration schedule, please refer to the event link.