Hi,
Is it possible to make computer vision projects with arduino or I need to use raspberry pi? If so can you tell me somewhere to learn it?
Thanks,
Machine learning with a 32 bit MCU is possible with the Nano 33 and the ESP32.
It is possible to use computer vision with some of the 32 bit microcontrollers and either TensorFlow Lite or Linear Regression.
The words TensorFlow and Linear Regression are keys that you can use to learn.
Use a Raspberry Pi. I have a ESP32-CAM that I am working with. The ESP32 CAM's image is being processed by the RPi. On the RPi I am using LineaRegression to ID trained models, "object recognition".
Using a ESP32 to process Linear Regression on estimating the next hours air pressure took 96 hours to develop a trained model. It takes my RPi moments to generate a trained model and guess hourly air pressure readings for the next 5 days. I been getting correlations of +.98xxx, with the RPi and the ESP32's.
Python is a great programing language and comes with many tools for ML, I am using, currently, Numpy and SCIKitLearn for linear regression (LR). I am still studying up on TensorFlow Lite.
There are several project using the ESM32-CAM that can do facial recognition.
The learning curve for TensorFlow Lite, for me, is a bit more steep, than linear regression. TensorFlow Lite will be a bit more accurate than LR. LR is easier to get working and does a pretty good job with images and trained models converted to grayscale.
The depository for TensorFlow is large and after making a contribution (your own Tensor) to the TensorFlow Lite community (Google), you'll get access to nearly all the Tensors that have been trained by Googles tools.
A laptop is great for learning computer vision. Plenty of memory, processor power and a camera!
My advice is get OpenCV and whatever hardware is needed to run it - I think the latest versions benefit enormously from the right kind of GPU being available. That strongly suggests full blown computer not microcontoller.
This topic was automatically closed 120 days after the last reply. New replies are no longer allowed.