I have zero knowledge about arduino tools, so can you suggest me what are the best components: arduino version (uno, micro), sensor, camera, motor wheel to use for this project?
Also worth to mention I will be using visual computation with python CV so perhaps I also need a strong cpu arduino?
have a look at ESP32-CAM
if is supported by the Arduino IDE
Is the project as specifified in the text of the image? Are you designing it for the Hecht Museum? Is it assumed that there will be a single visitor viewing a painting? From a relatively small, well defined location?
I will be implementing the project on a small scale (hand size objects)
I would look at this as a sensor:-
You can teach it various things like generic face recognition or specific face recognition. Then it communicates the information it sees to a lowly Arduino like a UNO through serial or I2C messages.
You can get them from UK distributors like Farnell.
And you won't need the openCV at all.
If that's a requirement, I'd look in the direction of a Raspberry instead of an Arduino.
Can you explain why based on your knowledge?
Well, there are many shades of grey and nothing would stop you from getting a powerful Arduino board and running python on it. The worlds of Arduino (microcontroller boards) and Raspberry (single board computers; SBC's) have come to overlap and in your case, this means you could go either way.
However, approaching this from where they have evolved from, I associate Arduino's with generally lower-weight/simpler computing tasks and lower-level programming, whereas a Raspberry -type board is aimed at higher-level languages and more computationally intensive tasks. If you're capturing video data and performing machine vision tasks on it, this is something that fits naturally in the world of SBC's, but for Arduino/microcontroller terms it's extending the top end of the range. Again, it's perfectly possible, but I personally don't regard it as the most obvious choice.
One practical argument is that in a typical Arduino setup, you'll spend a fair amount of work/energy on interfacing with the hardware and performing low-level operations in a language like C++. On a typical Raspberry/SBC, you would typically install an operating system and whatever device driver is necessary for your camera system and then proceed with the high-level application creation in something like Python. The latter is less efficient in terms of computing power, but tends to better fit a rapid-prototyping environment where very direct control over the hardware as such is not the main added value.
Since you seem to be approaching this much more from an applications-point of view and you imply a preference for Python, I would be inclined to nudge you in the direction of an SBC instead of a microcontroller.
What do you think about ESP32 or ESP32-CAM for python ai tasks compared to raspberry pi?
try a web search for esp-cam micropython you will get plenty of links which should give you some idea of its capabilities
clearly the Raspberry Pi 4 is more powerful than the ESP32 - it all depends on your requirements specification
I'd go with the Pi. And that comes from someone who's far more used to writing C++ on a microcontroller platform than Python on a higher level system!
which Pi version do you recommend, they so damn expensive compared to ESP32 lol
Sorry, can't really help you with that. You'd have to make a choice based on the computing power you need and that will depend on your specific application. Maybe someone else can chime in on this.
I do agree those ESP's are cheap as chips, but the savings in money may (depending on your knowledge and skills) not outweigh the time spent on getting the thing to do what you want. I expect you'll be up and running quicker with a Pi solution especially if you have no/little existing experience with embedded systems.
I haven't used these before but I'm intrigued by the marketing and price point for Hallowe'en applications. Prior to something like this I have used Arduino with an Xbox360 camera for skeleton tracking and crude gesture recognition, but that requires I also run a laptop with Windows 7 or 8 at latest. Certainly don't recommend this approach today.
Have you used the Huskylens and if so, did you find it easy to use every feature, or did it sort of do one thing well and the others so so?
Yes I have used it. There are lots more informative videos about that you get on that page I linked to. You should also search for the tutorials and manual for this device.
I showed it to some people at a disabled music group and they were impressed. One said it performs much better than the other camera he had, which was so useless he wished he had not bought it. He was referring to :-
https://pixycam.com/
I was mainly using the tag reading functions, although the generic face detection worked impressively well. That is finding a face without having to train it on what face you had to detect, which is the specific face recognition which I referred to. I have not tried the line following mode, but I have used the colour detection mode.
If you are into Lego, then they do a special version of the Huskylens, that you can program in a EV3 Lego system, using the same sort of user interface as that does. That is using blocks of commands rather like the Scratch language on the Pi.
Yes I have used an Xbox360 as well. I am sure that I got it to work by connecting it up to a Raspberry Pi. But the Raspberry Pi operating system has changed so much in the last year I am not confident it will still be able to run on one. I know that I used to be able to run the Processing language on the Raspberry Pi but now it will not work correctly.
Good to know, might come in useful.
I also tried on a RaspPi 4. Similarly, I wasn't able to get Processing to cooperate easily but heck if I can remember why. As for the Kinect, iirc, the trying to install the drivers would fail as I think the open Kinect drivers by Primesense (if that sounds familiar) were purchased by Apple or something like that. I still had the drivers for Windows 7/8 on an old laptop from 2015 or so that I was able to use. The Kinect works well enough for my purposes but certainly not something that would be reliable enough for the assistive devices.
Thanks for the details on the Huskylens.
This topic was automatically closed 180 days after the last reply. New replies are no longer allowed.
