Developers of ADAS systems have some major problems coping with their data and testing it.

Why is this Difficult?

  • Petabytes of streaming data in different locations
  • Hadoop, Spark, and traditional big data tools inadequate, too expensive and time consuming to centralize data
  • Neural networks are black boxes and you cannot test particular situations

“Machine learning systems, and deep learning in particular, are very high performance but don’t actually explain how they figure out the answer”
Gill Pratt, CEO Toyota Research

XCube Enables Efficient Deep Learning & True Scalability

  • Transforms problem of distributed data into the advantage of parallel processing
  • Automates training and testing of neural networks in parallel
  • Allows granular testing of situations, such as rain, snow, or four-way intersections
  • Infinitely superior to Hadoop or Spark for independent distributed data

By Working Where the Data Is

  • Use any desktop application, such as an automotive simulator
    • No need to rewrite it to run in parallel
  • Pick the data to run it on
    • Usually the data are the selected from previous searches
  • DSSC manages the parallel execution
    • Sends the application to where the data resides
    • Runs in parallel in virtual machines
    • Reformats the data to what the application needs
    • Collects the results

Compatible with Everything

  • Vendor independent Framework Support
    • Supports most popular frameworks natively (Caffe, Torch, Tensor Flow, Matlab)
  • Vendor Independent Hardware support
    • NVidia, Intel, Xeon-Phi, and new FPGA-CPU and other architectures
  • Used in small systems to global scale
    • No workflow change