Scene Segmentation Reference Design

CNN accelerated scene segmentation using Latticeā€™s Low Power FPGA

As edge devices become more and more aware of users and context of usage, devices such as PC, peripherals and personal assistance devices with video calling capability can implement intelligent user segmentation and background bluring for more private and collaborative experiences.

Low Power Accelerated Scene Segmentation – This reference design provides an implementation of scene segmentation on low power high performance Lattice FPGA.

Hardware Optimized Examples – complete design examples running on popular Lattice development boards (including FPGA RTL) include NN Models and sample training datasets and scripts to recreate and update the example as needed. Same examples are provided through sensAI Studio.

Features

  • Fast prototyping of solution and transfer learning to create additional use cases
  • Reference design includes all components needed to replicate the design
  • The entire end to end application implemented on FPGA resources including preprocessing of the video stream, scene segmentation using CNN based NN model and the background blurring

Block Diagram

Object Counting Reference Design Block Diagram using CrosLink-NX

Documentation

Technical Resources
TITLE NUMBER VERSION DATE FORMAT SIZE
Select All
CrossLink-NX Scene Segmentation Reference Design - Source Code
FPGA-RD-02256 1.0 6/7/2022 ZIP 32.3 MB
CrossLink-NX Scene Segmentation Reference Design - Documentation
FPGA-RD-02256 1.0 6/7/2022 PDF 2.7 MB

*By clicking on the "Notify Me of Changes" button, you agree to receive notifications on changes to the document(s) you selected.