MuseBox

FPGA Machine Learning based system for real-time AV Broadcasting applications

MuseBox is a real-time machine learning system designed for real-time AV Broadcasting applications. 

It can work with live stream, for interactive or live applications, and it can work with local files, when you have a big amount of files to process and also when these files are not accessible outside the local network for legal reasons.

It is based on a ZynQ Xilinx FPGA Stack for the multimedia elaboration and for Machine Learning elaboration, also with ALVEO Xilinx technology for On-Premise elaboration.

You can also connect the core system to your  industrial system through a standard protocol like ZMQ, RabbitMQ or ROS protocol.

MuseBox supports Vitis AI 1.0+ .

MuseBox machine learning on FPGA

Musebox actually supports different applications (>30FPS, realtime):

FACE ANALYSIS

  • Face detection
  • Face recognition (no mask) DEMO HERE
  • Face recognition (with mask) DEMO HERE
  • Face landmarking
  • Glasses detection
  • Mask detection
  • Gender detection
  • Eyes detection and blinking

 

PEOPLE ANALYSIS

 

AUDIO ANALYSIS

  • Audio clustering
  • Audio filtering
  • Audio segmentation

 

OBJECT ANALYSIS

  • Object detection
  • Object recognition
 

MEDICAL ANALYSIS

  • Medical segmentation

Actually MuseBox is supported in different hardware devices:

PYNQ Z2 (Zynq 7000)

 

  • Extra low power
  • Extra low cost
  • Embedded solution

ZCU104 (Zynq Ultrascale+)

  • Low power
  • Video optimized
  • Embedded solution

ALVEO U200

  • Mid power (200W)
  • Max performances
  • Edge solution (server)
  • product-ready on Server

ALVEO U50

  • Low power (75W)
  • High performance
  • Edge solution (server)
  • product-ready on Server

VERSAL VCK190