Meant to share this a while ago. So we have our initial online custom training for DepthAI now live on Colab.

https://colab.research.google.com/drive/1Eg-Pv7Amgc3THB6ZbnSaDJm0JAr0QPPU

So there are two notable limitations currently:

1. DepthAI currently supports OpenVINO 2019 R3, which itself requires older versions of TensorFlow and so on. So this flow has all those old versions, which causes a lot of additional steps in Colab... a lot of uninstalling current versions of stuff and installing old versions. We are currently in the process of upgrading our DepthAI codebase to support OpenVINO 2020.1, see here. The updated training flow when that's done. (EDIT: we got that done fast, see the bottom of this post for the 2020.1 training flow)
2. The final conversion for DepthAI (to .blob) for some reason will not run on Google Colab. So it requires a local machine to do it. We're planning on just making our own server for this purpose that Google Colab can talk to to do the conversion.

To test the custom training we took some images of apples and oranges and did a terrible job labeling them and then trained and converted the network and ran it on DepthAI. It's easy to get WAY better accuracies and detection rates by using something like basic.ai to generate a larger dataset.

To use the latest of everything (as of this writing), including OpenVINO (R2020.1), etc. use the following:

Training: https://colab.research.google.com/drive/1_bjLv6QH_SPQ4QQ4TX1l_45acBSbjQBu 

Running on DepthAI: https://github.com/luxonis/depthai-python-extras/tree/host_watchdog_r10.15

Cheers,

Brandon & the Luxonis Team

If anyone is in need of mounts for DepthAI, one of our active users made a whole set of enclosures and mounts for DepthAI variants:

• https://www.tinkercad.com/things/lUTygibtc12-bw1093-updated-single-camera
• https://www.tinkercad.com/things/iqgmT1IUlrS-luxonis-bk1093-single-camera-and-x
• https://www.tinkercad.com/things/4OlxITtchVX-luxonis-bw1098ffc-usb-3-edition
• https://www.tinkercad.com/things/79q1HPnB86X-luxonis-bw1097-rpi-case
• https://www.tinkercad.com/things/h1rb1lILS8P-luxonis-bw1098-obc-usb-3-with-integrated-cameras

Hi everyone,

Just wanted to share the the Crowd Supply shipments have been received, and our fans are up and running with the hardware, with much exciting activity and discussion in our Community Slack channel.

To join, go to our docs page: https://docs.luxonis.com/ and scroll down to the bottom and click on 'Slack Community' to join.

We also have a couple new hardware models for sale, and a PoE variant of DepthAI in progress.

First, the two new models since the Crowd Supply campaign (these will be on the campaign soon):

1. BW1098OBC - USB3, Onboard cameras:

2. BW1093 Single camera variant, USB3C.  We're calling this uAI (pronounced 'micro AI').

You can buy initial versions of these here:

https://shop.luxonis.com/

Note that the BW1093 has a temporary heatsink on these units, as the Coronavirus is preventing the CM from being able to make the final version.  So it's just a stick-on one for now, similar to what the Raspberry Pi stick-on heatsinks look like:

And for the new board we're working on:

We got requests to be able to deploy DepthAI quite far from any host computer.  And to be able to deploy as many as people want.  Boom, enter PoE.  Now you can deploy DepthAI at up to 328.1 feet (100 meters, for those who have 1m = 3.281 feet memorized, like I do, strangely).

Below are the renderings of this Myriad X power over Ethernet Design.  This is more or less our initial test design.  Other variants will come after this is validated and proves out all the EE pieces.  

Our plan is to make an ROS module so that when plugged into the network ROS will auto-discover it and it can be used right away.

Thoughts?

Thanks,

Brandon and the Luxonis team!

Hi all,

So 1/2 of the Crowd Supply orders shipped to customers yesterday.  And the other half are shipping out today.  

So guess what?  Our planned shipped date was... drum roll please, February 14th (today)!

And we have our MVP running:

So what is this showing?  The dawn of 'spatial AI'.  Where now your embedded system can know what objects are, and where they are in physical space, in real time.  

It's a capability that straight-up didn't exist at all prior to 2016, even on mainframes.  And even recently, thinking you could do this on an embedded system would be considered science fiction.

Now you can have this capability, for your product, with this little module:

And big thanks to the Crowd Supply team for getting these all shipped out on time (and early)!

Best,

Brandon & the Luxonis team.

Hi DepthAI fans!

So the following PCBAs just finished soldered assembly (final assembly is still to be done):
 - DepthAI: USB3 Edition
 - DepthAI: System on Module
 - 4K, 60 Hz Video Modular Color Camera
 
 Below are some pictures of the assembled/soldered boards and them shipping out to the US.
 **DepthAI: USB3 Edition**

 
 **DepthAI: System on Module*

 
 **4K, 60 Hz Video Modular Color Camera**

 
 The following DepthAI variants are undergoing soldered assembly now, with the goal to finish before Chinese New Year starts.
 - DepthAI: RPi Compute Module Edition
 - DepthAI: RPi HAT Edition
 - 720p, 120 Hz Global Shutter Modular Stereo Camera Pair
 
So this year Chinese New Year is January 18th through February 2nd, where most companies will be offline.  The remainder
will be offline January 11th through the February 2nd.  

So our strategy on fulfillment is (and has been) to conclude all the China-involved production prior to January 11th, doing the remainder of the full-system assembly and test in the United States while China is offline to hit our target of DepthAI CrowdSupply shipments by February 14th (to show our love!).  

It looks like we're still on schedule to hit this, but it going to be close!

We'll keep everyone updated.  

Thanks,
Brandon and the Luxonis Team

Hey everyone,

So out of the model variants we've made/released on our Crowd Supply (here), we've gotten a TON of feedback to have two other variants:

1. Single-camera, USB interface, and TINY.  (So we made and ordered that, see last update here)
2. Tri-Camera, USB interface, and human-eye camera spacing.  Allowing easy plug/play onto platforms where the human-eye spacing is sufficient for the stereo baseline.

So, we made 2 as well, and it's arrived!, and works great!  Pictures below.

Here's our first check of it (before thermal-padding/gluing the cameras down), just to make sure all the cameras work as intended.  This is also of course not calibrate for depth (as the cameras are flowing by their FFC connections at this point).

Thoughts?

Cheers,

Brandon & the Luxonis Team

Hey Guys,

So one piece of feedback we've gotten a ton of is 'That whole Depth + AI thing is great, and we're going to use it on several efforts... but, do you have a small single-camera version for AI only without depth?'

So, we're making it.  And we're making it small:

So it looks like we're finish the simplified single camera design today (heavily leveraging our existing module) and likely get it ordered tomorrow.  So this means we'll have it in-hand and ready to go probably mid December.

^ The gray box is the color camera (we haven't bothered to take the time to make a model of it).  

And the pink part is the thermal interface material (TIM) which connects the Myriad X to the heatsink.  It's the tallest part on the back, intentionally, to make it so this can be mounted easily to existing hardware and heat-sink to it.

To put a scale to this, it's approximately the same dimensions as this (the unit above is 44mm x 30mm, and the module below is 40mm x 30mm):
And if anyone is interest in this, we do have an initial version that our intern made, which is a two-board solution (which is much larger... hey, it's an intern project!) but is a working system:

Cheers,

Brandon and the Luxonis Team!

Hey everyone,

Our Crowd Supply campaign is live!  Please help us bring DepthAI to the world by backing our campaign!

https://www.crowdsupply.com/luxonis/depthai

Thanks!

Brandon and the Luxonis Team

Hi everyone,

Our Crowd Supply campaign is going live on Monday!

So if you haven’t done so already, please subscribe on our project page to get notified when the DepthAI campaign starts!

https://www.crowdsupply.com/luxonis/depthai

Best,

Brandon and the Luxonis Team!

Hi everyone,

So we've been heads-down preparing for our Crowd Supply campaign launch (hence a bit of lack of updates here lately). 

In preparation for that, we got our github up, which should the campaign succeed, will host our source design files for both hardware and software.  (If the campaign doesn't succeed, we're going to keep these hardware and software source files internal for our own use only).

So without further ado, below are some details on the Crowd Supply campaign (which can be found on our github as well, here).

DepthAI Crowd Supply Campaign

We are launching a CrowdSupply campaign (https://www.crowdsupply.com/luxonis/depthai) to bring the power of the Myriad X to your design. If the campaign is successful we will be releasing the designs (hardware and software) for the following Myriad X carrier boards here on Github - to enable engineers and makers to build their own solutions, leveraging our source designs/code.

In addition to the designs shown above (which will be purchase-able through the campaign), we will be releasing an ordered, but not-yet-received version which integrates all 3 cameras onto a single board with a USB3C interface to the host, shown on the bottom right below.

The Why of DepthAI

The Myriad X is WAY faster and a TON of features are only accessible when cameras are connected directly to it over MIPI. Take object detection below as an example: 

So that's 25FPS instead of 8FPS... and that frame-rate is achieved with no host processing AT ALL! (Technically, you can run DepthAI with no host at all - but often you'll want a host so that other business-logic code can be run based on the output of DepthAI... like 'pick only the ripe strawberries')

There's more to it than just being faster though... the DepthAI platform is engineered from the ground up to enable the original vision of the Myriad X - which is to allow low-power, high-efficiency vision processing including stereo depth, motion estimation, and neural inference. Existing platforms have no MIPI interface and only have a PCIE or USB inferface and so cannot take advantage of a TON of the hardware modules built into the Myriad X which are only (meaningfully) usable with MIPI:

• Stereo Depth
• Edge Detection
• Harris Filtering
• Warp/De-Warp
• Motion Estimation
• ISP Pipeline
• JPEG Encoding
• H.264 and H.265 Encoding

The How and What of DepthAI:

So to allow the use of these, DepthAI is engineered to allow direct MIPI connection:

And to allow this power to be integrate-able into actual products (which may require differing stereo baselines, differing interfaces, geometries, etc.) and to make hardware integration much easier (removing the need to integrate a very-fine-pitch multi-hundred-ball BGA into your design) we made a System on Module to hold the Myriad X:

This module enables a simple and standard tolerance PCB, instead of the high-layer-count and high-density-integration (including laser vias and stacked vias) required to integrate the Myriad X directly. And it also provides all the power supply rails, power sequencing, clock generation, and camera-support hardware on-module.

This allows you to integrate the power of the Myriad X into your design with a standard 4-layer (or even 2-layer, gasp!) PCB. And then leverage our reference hardware and software designs to get up/going super-fast (or just buy our boards directly, if they happen to fit the bill for what you need). Which all results in getting up/running fast: 

So what are the reference designs (the carrier boards) which will be released should the campaign be funded?

DepthAI | Raspberry Pi Compute Module

This design allows you to have the Myriad X, all the cameras, and the Raspberry Pi on one board. So that you don't have to worry about cabling things together, boards strewn about your desk, etc. when prototyping and writing code.

DepthAI | Raspberry Pi HAT with FFC Camera Interfaces

This design allows you to leverage your existing Raspberry Pi (and perhaps even its mounting on your platform) and simply add this DepthAI hat to get this AI, depth, and vision processing power. Off-board modular cameras allow integration up to 6 inches away from the Pi/HAT.

DepthAI | USB3 with FFC Camera Interfaces

This design allows you to leverage DepthAI with whichever platform you choose. It works with anything that runs OpenVINO, which is a lot of systems.

• Linux: Ubuntu - Yocto Project - CentOS - Raspbian
• Mac OS X
• Windows 10

Note: We have a preliminary system working which actually works with the Raspberry Pi Camera V2.1... but the ISP is still a bit broken so it flashes at lighting transitions/etc. If there's sufficient interest we'll finish this proof-of-concept driver/camera tuning for the Myriad X and release hardware to support the Pi V2.1 camera as well.

DepthAI | Modular (FFC) Cameras

We designed these modular cameras to have the same mounting pattern and board size as the Raspberry Pi V2.1 camera so that existing mounts/etc. likely can be used (or lightly modified and used). That and, the size was going to be super close anyways, so why not go ahead and -not- make yet another mechanical mounting standard?

Want to build things off this? Or want to buy these straight up? Support our CrowdSupply to make this a reality: https://www.crowdsupply.com/luxonis/depthai

Best, 

Brandon and the Luxonis team!