Now the the times come to gain knowledge by doing experiment for machine learning project. I have a raspberry Pi3 and Pi camera. Please suggest me useful machine learning project that i can test with pi and pi camera.

Would fire detection be a good example for a machine learning project?
@nsaspook @xox

 

What do you think about predicting temperature from humidity. is this irrelevant example?

 

What do you think about predicting temperature from humidity. is this irrelevant example?

Not relevant and not possible.
The two are not related. You can have temperature at 0°C and humidity at 0-100%. Similarly you can have temperature at 35°C and humidity at 0-100%.

Temperature, humidity, and dewpoint are interrelated.

 

couple voice recognition with facial recognition.

 

couple voice recognition with facial recognition.

i have Pi camera for facial recognition but i have nothing for voice recognition. Maybe I'll need mike and voice recognition ic for this that I don't have, That's why I am looking something simple to perform experiment

 

Figure out a bird or insect species recognition system based on camera.
Or leaf or flower recognition.

 

Figure out a bird or insect species recognition system based on camera.
Or leaf or flower recognition.

So to make this type of system, we will need many images. We have to create a mathematical-model in which these images will be fed as data input. I have seen that machine learning requires GPU. Is the Pi capable enough to train the model?

 

So to make this type of system, we will need many images. We have to create a mathematical-model in which these images will be fed as data input. I have seen that machine learning requires GPU. Is the Pi capable enough to train the model?

Yes, if you don't expect an extremely fast interface. Also, there is a limited number of bird species in each country vs doing facial recognition and looking for a single person out of a country full of people.

if you need more speed and still have to use a pi, then you can use the Pi to control the interface and camera - and you can feed the images to an AWS account that processes the images with an array of GPUs and manages the database on an elastic compute instance. The limit is up to your imagination and budget.

 

if you need more speed and still have to use a pi, then you can use the Pi to control the interface and camera - and you can feed the images to an AWS account that processes the images with an array of GPUs and manages the database on an elastic compute instance. The limit is up to your imagination and budget.

please correct me if I am wrong in my thinking. i think i have to write two programs. The first will be for Pi and the second will be for AWS Cloud.

I need to write a Pi program to capture the image and send it to the AWS cloud.

The model will run in AWS, so where do I have to write the program for this model, do I have to write it on PC and deploy to AWS cloud?

 

please correct me if I am wrong in my thinking. i think i have to write two programs. The first will be for Pi and the second will be for AWS Cloud.

I need to write a Pi program to capture the image and send it to the AWS cloud.

The model will run in AWS, so where do I have to write the program for this model, do I have to write it on PC and deploy to AWS cloud?

Start by doing everything on the Pi.
Only migrate to AWS if you need more speed. Or if you want to set up web access to everyone. Setting up an AWS account is a project in itself