Hi DepthAI Fans,

We’re continuing to work hard on DepthAI resources, so we have an exciting slew of updates for you!

Custom Training Tutorial

We now have a Custom Training Tutorial which uses Google Colab. So you can train your own custom models for DepthAI completely for free.

• https://docs.luxonis.com/tutorials/object_det_mnssv2_training/

Other Progress

Additionally, since the last update, we:

• Improved stereo depth synchronization - improving overall depth quality - and in particular improving quality under fast motion.
• Enabled a USB2 only mode to allow use of longer cables and resolve issues some Linux distros and hardware permutations have with USB3
• Updated our firmware and software to support OpenVINO 2020.1.
• Released an open-source regression-testing script (which can help users see various use-cases of the DepthAI API
• Implemented support for Mac OS X / HomeBrew (which is awesome, it's my new favorite feature).
• Implemented a C++-level watchdog and monitors and recovers should something go wrong w/ the device (say if someone hits the reset button on the board during your big demo, the system will just recover without your code crashing - it will just see a delay while it recovers).
• Implemented a host-side supervisor which automatically makes DepthAI recover from exceptions at the Python level. Accidentally don't catch a NaN going into a plot? No problem, DephtAI will do that for you.
• Open Sourced the C++ DepthAI API. This enables pip install support directly, which we'll be working on.
• Fixed a USB3 issue with the Raspberry Pi 4. Now you can happily use full USB3 speed with the Raspberry Pi 4.
• Published a bunch of documentation (relatively speaking) on docs.luxonis.com

There’s probably more, but the list was getting long!

And the best is yet to come. Even with all the progress above, there are a bunch of other features that we’re excited to be working on, and will share with you when we can.

Stay safe and best wishes,
Brandon and the Luxonis team