Product FAQ

[idata]

Q: Where can I buy the Neural Compute Stick?

A: Please visit https://developer.movidius.com/buy

Q: The Neural Compute Stick is not available in my country, so how can I buy it?

A: We are constantly expanding our global distribution. As more countries are added, we will update the list of global distributors at: https://developer.movidius.com/buy

Q: What are some of the key features of Neural Compute Stick?

A: The Movidius™ Neural Compute stick is a development and prototyping platform for machine intelligence at the edge:

• Profile, tune and compile your own neural networks for embedded applications
• Add machine intelligence capability to Linux systems with a USB Type-A port
• Develop and prototype with existing CNN models or use your custom trained networks
• Become a developer for the low-power high-performance Movidius™ VPU - a discrete VPU which is used in millions of smart security cameras, gesture controlled drones, industrial machine vision equipment, etc.

Q: What kind of support can I get for the NCS?

A: https://developer.movidius.com is an active and vibrant developer community hosted by the makers of Movidius™ Neural Compute stick. You can collaborate with fellow deep-learning developers, ask troubleshooting questions, share your thoughts & projects on deep-learning inference, etc. Register at https://developer.movidius.com/forums to be part of this community. Developer blogs detailing deployment and SDK tools can be found at https://movidius.github.io/blog.

Q: What documentation is available for Neural Compute Stick?

A: Please visit https://developer.movidius.com/start for a quick getting started guide. For detailed documentation on the SDK and the API as well as release notes, visit https://movidius.github.io/ncsdk/ and https://github.com/movidius/ncsdk/blob/master/docs/release_notes.md.

Q: What developer tools are available for NCS?

A: The Movidius™ Neural Compute Stick comes with a freely downloadable software developer kit called the Movidius™ Neural Compute SDK. Please visit https://developer.movidius.com/start for more information. To download and install the SDK, please visit https://github.com/movidius/ncsdk.

Q: What is the Neural Compute SDK?

A: The Movidius™ Neural Compute SDK is a collection of developer tools which can be used to develop machine intelligence inference applications for NCS.

Q: What platforms and Operating Systems does the Neural Compute Stick work on?

A: Ubuntu 16.04 installed on a physical x86_64 system, Stretch running on Raspberry PI 3 Model B and Ubuntu 16.04 VirtualBox instance running on Windows 10 or OSX. Installations on other OS may be possible using an API only installation and installing all required dependencies. An example of API-only installation can be found at https://movidius.github.io/blog/ncs-apps-on-rpi/.

Q: Does the Neural Compute Stick work on my Windows machine?

A: The Movidius™ Neural Compute SDK can be installed on Ubuntu 16.04 VirtualBox instance running on a Windows 10 machine.

Q: Which networks does the Neural Compute Stick support?

A: Please visit the SDK release notes for a full list of supported networks.

Q: Will the MvNC SDK work with Fathom Neural Compute Stick?

A: MvNC SDK is designed to work only with the Movidius Neural Compute Stick, we haven't done any compatibility testing with the Fathom Neural Compute Stick. Please visit https://developer.movidius.com/buy to buy your Movidius NCS.

Q: Can I use an NCS device for training?

A: The NCS device is designed for deploying neural networks and does not play a role in the training portion of the neural network. It's main role is to act as a co-processor for processing inferences.

Q: What programming languages are available in the Movdius™ Neural Compute SDK?

A: The Neural Compute SDK comes with a C++ and Python API. (working with Python 2.7 and 3.5).

[idata]

Q: How much memory does the Movidius™ Neural Compute Stick have?

A: The NCS device has 4 Gb (500 MB) of memory that is used for intermediate processing and allocation of model graph files.