• DepthAI-v2

  • Modular Design Ordered (Including a Test Board)

Hi everyone!

So we ordered the Modular design for the Movidius Myriad X. As below it breaks out MIPI, and will be the core of our products.

BUT! It also allows you to integrate this into your prototypes and/or products. Without the hassle of dealing w/ a super-fine-pitch tiny 400 pin BGA!

Instead it's an easy Hirose connector that gives you everyone you want, and nothing more. In a convenient 40mm x 30mm package.

We do have a question:

Should we make one of these with a built-in host controller, say an i.MX 8M?

And oh by the way here's the breakout board, which allows connection to the IMX378 modular sensor and also 2x OV09282 (for stereo disparity depth):

Cheers,
The Luxonis Team

2 months later

Hello I am Bob from Kenya a software innovator with Intel . I have the movidius stick and I love what you guys are doing .In fact I am working on a solution using movidius myriad x chip but the designs has been hectic and crasy so please I would prefer you keep the SOM without the imx 8 microprocessor controller . would you be kind eneough to ship for me 2 such dev kits the som module and the break out board .how much would it cost for you to send them to me in Nairobi via DHL thanks ..good work man

Hi BobAfwata!

Yes, we can definitely get you two kits. And yes we have the SoM w/ Myriad X only in-hand and it's working. And we do intend to sell it as a standard product. Right now we're waiting on heatsinks so that it can be used on a bench for development. (For final products it can just be heatsinked to the product enclosure instead.)

Best,
Brandon

BobAfwata Yes we'd be able to make it in that form-factor. Our current modular solution seems to be close to that already.

Do you need that exactly? Or are you just looking for something that's small?

Thanks,
Brandon

    10 days later

    Have a look at this looks like competition

    4 days later

    So yes we have one of those. So actually that's not competition. So our system's value add to the world is to provide real-time object localization in 3D space. So this requires around 2 tera operations per second of calculations (2 TOPS). So that product is, roughly speaking, capable of (round numbers) 50 million operations per second (50 MOPS).

    So DepthAI is about, round numbers, 500,000 time faster than that solution. But really they shouldn't be compared... that solution is for embedded voice recognition. (Sparkfun is working on supporting a camera, but it's still TBD, and at least for OpenMV, for running something like MobileNet, external DRAM is required - so it could prove to not be possible to run camera ML models on this device.)

    It's just a different category of embedded, as well. So the purpose of that device is to be able to run on a coin-cell battery for years. Our platform is intended to run on a 7Ah battery for 10 hours. :-)

    So in terms of devices that can run image/video neural models, see here: https://discuss.luxonis.com/d/4-neural-processors-and-hosts-we-ve-found-so-far

    And in particular, check out Greenwaves if you're looking for super-low-power neural inference from video/image sensors. As if you want pure-embedded and very low power, this is a good solution:

    Thoughts?

    Thanks,
    Brandon

    a month later

    Interesting. Thanks for pointing things like this out. Yes, this is the traditional use of the NCS2 where the host has to to do all the processing and the stereo depth engine/etc. in the Myriad X is not useable.

    And also as a result the power use is WAY higher.

    So for a power-use comparison to systems which are already much lower than this, check out here:
    https://hackaday.io/project/163679-luxonis-depthai/log/169487-power-use-comparison

    Thoughts?

    Best,
    Brandon

    a month later