What is PetaVision?

PetaVision is an open source, object oriented neural simulation toolbox optimized for high-performance multi-core, multi-node computer architectures. PetaVision is intended for computational neuroscientists who seek to apply neuromorphic models to hard signal processing problems; both to improve on the performance of existing algorithms and/or to gain insight into the computational mechanisms underlying biological neural processing.

Core PetaVision Features:

  • Open-sourced + developer friendly
  • C / C++ for fast execution and OOP
  • Optimized for parallelization (OpenMP, MPI, cuDNN)
  • Based on local connectivity
  • Experiments defined through a text file

How has PetaVision been used?

Petavision has been used on a range of vision and learning tasks including depth inference, action classification, and learning from multiple intelligence streams including weather and commodity prices. PetaVision has been used to perform image classification, sound analysis, and image-rescaling and is flexible enough to grow with your imagination.

Getting Started

We recommend that you jump over to our growing wiki page on SourceForge to find installation instructions, tutorials, and more about how PetaVision works and how it can work for you.

Our code is documented using doxygen and you can explore the code in more detail, please refer to our documentation.

Try PetaVision on the Amazon Cloud

If you have an account on Amazon Web Services, you can try PetaVision using our public AMI on aws.amazon.com. Search for "PetaVision Public AMI" and start an instance. Use a GPU instance (e.g. g2.8xlarge) to take full advantage of PetaVision's acceleration toolkit. Follow the instructions in our documentation for AWS to get started.

PetaVision: A Guided Tour (April 24, 2015)

Follow along here with a video summary of PetaVision