I have my arm almost together now

I have started working on a self-driving project> Check it out here: http://bit.ly/SelfDriveV1

• What project do you have planned with the BeagleBone AI?

I'd like to try signal processing with the DSP on radio and audio signals in ham radio.

• What familiarity you have with programming and linux.

I've been programming since a kid with a Commodore 64. 'nix since before there was a www.

• If you've already done a project, how you plan to transfer it over to the BeagleBone AI

Example: https://hackaday.com/2019/01/05/google-assistant-now-available-on-ham-radio/

I've worked with the BeagleBone Black boards already. They're great and have plenty of GPIO.

I'm starting to think vision in that would be more fun than in the moose. The moose is still getting done as an animatronic assistant either way, but the twilight zone fortune telling with vision to both get mood and recognize people, topped off with pulling random "real" astrology off the web would be a lot of fun. There seems to be quite a few astrology APIs out there

https://www.programmableweb.com/category/astrology/api

I personally have not reid other capes yet. according to the Posts regarding these, they are currently working on compability, the device tree overlays have to be rewritten to make them compatible.
I have high hopes that they will resolve compatibility issues pretty fast.

On LTE and MCU with LTE connctivity might be lighter than the usb stick but for sure the stick with existing linux drivers might be easier to use. Be aware that the Debian version that runsd on the BB-AI is not vanilla debian therefore the drivers might not be in the repos but have to be manually isntalled.

mayermakes , thank you for your insight. Have you ever tried on interfacing any other capes? I came to know from the documentation that BB-AI would be compatible with the existing BBB capes.

Regarding LTE, since i want to implement this on drone, adding an extra companion computer would seem not viable as there are weight constraints. My plan was to interface a huawei LTE usb modem since their drivers are compatible with linux.

Thanks for sharing information on getting started. In the community I have seen that its dificult to access the PRUs, etc. Your projects and tutorials will help me a lot!

In the recent years ML especially image processing has taken a huge leap. Amazon can suggest a product based on picture. Facebook automatically tags your friends. With its recent advancements I think detecting 4-5 plant diseases is possible. It will be a fun ride!

Yes, I have a PocketBeagle and have used it quite a few times.

So there is a challenge with the moose puppet now that I have it. It is real moose head size, which is great, but its purposefully overstuffed so that it holds strongly onto your arm when you go to move the mouth. So to give it good animatronics for motion, i'll have to open it up, remove all the stuffing (or most) and create some sort of flexible internal skeleton to allow the whole head to move. I may take that as a "phase 2" item, and just concentrate on the mouth movement syncing with the spoken google assistant response.

Vision would be fun to work in along with an AI. having him recognize people, and track their most frequent requests, and asking them if they want to "check the sunset time" or "when is high tide" or whatever when they walk past would be interesting.

woah you got quite an ambitious project in front of you! here is some thing that may help you i ngetting started:
https://beagleboard.org/p/patrickpoirier51/pocketpilot-an-autopilot-based-on-the-25-pocketbeagle-1fa14b

this autopilot concept for the pocketbeagle also runs on the BBB so you might be able to port it to the BBAI- be aware of the ongoing findings regarding the GPIO/PRU pins ant their mapping!

it should givbe you a point to start for navigation

there are also other projects that aim to  identify bugs with the BBAI so collaboration with those is advised!

ON detecting plant deseases:

My old day job was crop protection specialist and part of that was identifying deseases for farmers and private people. it is not doable without looking very closely and knowing some quirks of a lot of deseases(for example sporulation conditions and inkubation times in certain areas regarding current weather conditions and even the exact field the probe was taken from- I really asked winegrowers about the exact locations and recalled similar findings from past years and known quirks like tendencies for moisture to fade slower than on the field next to that location). all of this stuff goes into a correct identification and treatment suggestion. i ALWAYS looked at the probes with a magnifying glass, and I would not trust anybody to be able to identify all the possible causes without it. a LOT of experience is involved in getting it correct and nobody is without fault in that field.
So doing that via image recognition is a HUGE feat. if you manage to get the MAIN 3-4 deseases with typical full blown outbreaks for ONE particular crop identified correctly that alone would be a tremendous success.

Do not let these insights deter you, go for it and try it out you may find identificator keys in non visible spectra that a human could not identify per eye, but AI could.

Wish you the best of luck but even more patience and persistence with that project!