To see product details, add this item to your cart. You can always remove it later.
Added to

Sorry, there was a problem.

There was an error retrieving your Wish Lists. Please try again.

Sorry, there was a problem.

List unavailable.

Intel NCSM2450.DK1 Movidius Neural Compute Stick

3.8 3.8 out of 5 stars 43 ratings

To see product details, add this item to your cart. You can always remove it later.

Purchase options and add-ons

Brand Intel
Connectivity Technology USB
Included Components USB Stick
Wireless Communication Standard Bluetooth
Processor Count 1

About this item

  • Neural Network Accelerator in USB Stick Form Factor
  • Real-time on-device inference; no cloud connectivity required
  • No additional heat-sink, no fan, no cables, no additional power supply
  • Prototype, tune, validate and deploy deep neural networks at the edge
Serve a pot of gold
Get ready for St Patrick's Day Find a Fresh store

Frequently bought together

$59.00
Get it Mar 21 - 25
In Stock
Ships from and sold by OEM XS INC..
+
$61.75
Get it as soon as Sunday, Mar 24
In Stock
Sold by MemoryWhiz and ships from Amazon Fulfillment.
Total price:
To see our price, add these items to your cart.
Details
Added to Cart
These items are shipped from and sold by different sellers.
Choose items to buy together.

Important information

To report an issue with this product or seller, click here.

What's in the box

  • USB Stick
  • Looking for specific info?

    Product information

    Technical Details

    Collapse all

    Additional Information

    Warranty & Support

    Amazon.com Return Policy:You may return any new computer purchased from Amazon.com that is "dead on arrival," arrives in damaged condition, or is still in unopened boxes, for a full refund within 30 days of purchase. Amazon.com reserves the right to test "dead on arrival" returns and impose a customer fee equal to 15 percent of the product sales price if the customer misrepresents the condition of the product. Any returned computer that is damaged through customer misuse, is missing parts, or is in unsellable condition due to customer tampering will result in the customer being charged a higher restocking fee based on the condition of the product. Amazon.com will not accept returns of any desktop or notebook computer more than 30 days after you receive the shipment. New, used, and refurbished products purchased from Marketplace vendors are subject to the returns policy of the individual vendor.
    Product Warranty: For warranty information about this product, please click here

    Feedback

    Intel NCSM2450.DK1 Movidius Neural Compute Stick


    Found a lower price? Let us know. Although we can't match every price reported, we'll use your feedback to ensure that our prices remain competitive.

    Where did you see a lower price?

    URL:
    Price: ($)
    Shipping cost: ($)
    Date of the price:
    /
    /

    Store name:
    City:
    State:
    Price: ($)
    Date of the price:
    /
    /


    Please sign in to provide feedback.

    Product Description

    The Movidius Neural Compute Stick is a miniature deep learning hardware development platform that you can use to prototype, tune, and validate, your AI programs, specifically Deep Neural Networks. It features the same Movidius vision processing unit (VPU) used to bring machine intelligence to drones, surveillance cameras, and VR or AR headsets now, in a USB stick form factor.

    Customer reviews

    3.8 out of 5 stars
    3.8 out of 5
    43 global ratings

    Customers say

    Customers like the ease of installation and speed of the single board computer. They mention that the SDK is easy to install and the getting started guide looked simple. They also appreciate the performance boost and object detection is very fast. That said, some disagree on simplicity.

    AI-generated from the text of customer reviews

    3 customers mention3 positive0 negative

    Customers find the installation process of the single board computer to be easy. They mention that the SDK is easy to install and the getting started guide looks simple.

    "...The SDK is easy to install. I'm a beginner, and following pyimagesearch movidus tutorials, and its working great...." Read more

    "I really wanted to have fun with this. The getting started guide looked simple. The video looked cool. Then I bought it...." Read more

    "Install instruction did work. Worse it corrupted my opencv. Venedor hasn't responded through Amazon nor to contact on the website." Read more

    3 customers mention3 positive0 negative

    Customers are satisfied with the speed of the single board computer. They mention that it is very fast, it has a magical performance boost, and the object detection is really fast.

    "...2. Low power - it gets it's power from the USB port.3. Very fast - the examples that ship with it are well-known deep learning test cases and..." Read more

    "the object detection is realy fast with this module. (i use it with a raspberry pi V3 and 5 frame/sec.) i love it!" Read more

    "Magical performance boost..." Read more

    9 customers mention5 positive4 negative

    Customers have mixed opinions about the simplicity of the board. Some mention that it works great on their desktop, while others say that it doesn't work at all.

    "...a beginner, and following pyimagesearch movidus tutorials, and its working great. Need an extra computer for the SDK though." Read more

    "The bad news first:1. The device is only usable on a a Raspberry Pi or a PC running Ubuntu 16.04 LTS...." Read more

    "Worked great on my desktop connected to a VMware workstation virtual machine running Ubuntu 16.04 LTS...." Read more

    "Works as promised. Used with a Raspberry Pi 3. Not for a beginner. Using this unit requires expertise well beyond a beginner's knowledge." Read more

    Good next step for low power portable deep neural network deployment.
    5 Stars
    Good next step for low power portable deep neural network deployment.
    I bought mine through Mouser since Amazon didn’t have any available through Prime at the time.Several thoughts about this device:1. It’s for running a deep neural network... oriented towards convolutional neural nets. You load your neural network configuration and trained weights onto this module and feed it data to process.2. This device won’t do the calculations related to training of neural networks, so it can’t provide acceleration in that regard.3. You won’t get GPU level performance from this device. It will give you desktop multi-core CPU level compute ability, as it relates to DNN(s) for low power systems, but it won’t give you GPU level compute abilities.4. This is a USB2 and USB3 device. (See below)So the currently available libraries are basically targeting Ubuntu Linux machines, ie, it;’s targeting the Raspberrry Pi. However, you can run it on a physical Ubuntu box or a virtual Ubuntu box. You just download the Intel SDK for this as well as the examples, install the required dependencies, and compile the codes. Works great.To get it to work on a virtual machine, you need to do USB pass through. When the device is first detected, it shows up as a USB2 device. However, once the DNN code is loaded onto it, it transitions and changes into a USB 3 device. So your virtual machine needs to do pass through for both versions of the device. Basically, run the code to start it. The device will disappear, your code will wait. During this time, search for a new device that just showed up and add it to the pass through. Restart your DNN test program and it should see it and use it this time.Having multiple NCS sticks and using them all is definitely possible. However your code needs to recognize the fact that there is more than one and work with them appropriately.You will NOT be able to run a larger DNN across multiple NCS(s). At least not with the current SDK. You can do multi-staging, but not spanning. You can, however, run different situations on different NCS(s). I was able to run 4 at the same time on 4 different NCS modules. They get enumerated during discovery and you just have the code choose one by index number.I think this is a pretty good device and for the low power consumption, it is geared towards portable/mobile use cases and not for replacing GPU(s).As others have noted, this thing is wide. So you might have trouble getting them into a USB hub properly. I was able to get them all on one hub because the spacing of the USB ports was wider and I alternated between the NCS and normal plugs. (See photo)
    Thank you for your feedback
    Sorry, there was an error
    Sorry we couldn't load the review

    Top reviews from the United States

    Reviewed in the United States on November 12, 2017
    88 people found this helpful
    Report
    Reviewed in the United States on July 5, 2018
    5 people found this helpful
    Report
    Reviewed in the United States on December 14, 2017
    21 people found this helpful
    Report
    Reviewed in the United States on October 29, 2018
    6 people found this helpful
    Report
    Reviewed in the United States on November 18, 2018
    One person found this helpful
    Report
    Reviewed in the United States on March 28, 2018
    3 people found this helpful
    Report
    Reviewed in the United States on June 9, 2018
    8 people found this helpful
    Report
    Reviewed in the United States on August 1, 2018
    20 people found this helpful
    Report

    Top reviews from other countries

    Amazon Customer
    5.0 out of 5 stars Five Stars
    Reviewed in Canada on June 18, 2018