Has anyone used machine learning or AI in their a project?

Thread Starter

rt694157

Joined Dec 15, 2019
57
Hi

I have never had or need machine learning or AI yet. I am curious to know if anyone here has used machine learning or AI in their project.
Can anyone share their thoughts on when and why we should use machine learning or AI in project ?
 

KeithWalker

Joined Jul 10, 2017
1,154
I will not give you my thoughts on machine learning or AI as there is an immense amount of information readily available on the internet. However, I will share information on a simple machine learning project that I made:
I got tired of playing with a laser pointer for my cat so I made an automated one. I used an arduino mini and a couple of cheap servos. I realized the to be useful in different locations, I needed to be able to alter the movement pattern. I added a small joystick and a couple of buttons so that I could create different patterns.
If the SET button is pushed, it goes into learning mode. The PROGRAM light turns on. The joystick is used to position the red spot. A SAVE button saves the coordinates for the servos for that location. Each time the joystick is moved and the SAVE button is pressed, a new location is saved.
When the pattern is complete, pressing the SET button turns on the PLAYTIME light. Each time the SAVE button is pressed, the number of times the pattern will be repeated is incremented by one.
Pressing the SET button then turns on the TIMEOUT light. Each time the SAVE button is pressed it increments the off interval between patterns by five minutes.
Finally pressing the SET button turns off the program lights and the toy starts playing the pattern.
The learned information on coordinates, repeat times and off interval are stored in non-volatile memory so that when it is powered off and then on, the saved pattern will start playing.
Keith

CatToy_2.jpgCatToy_1.jpg
 
Last edited:

nsaspook

Joined Aug 27, 2009
7,494
Hi

I have never had or need machine learning or AI yet. I am curious to know if anyone here has used machine learning or AI in their project.
Can anyone share their thoughts on when and why we should use machine learning or AI in project ?
Sure, most of us have designed and used a Expert System is a engineering type project. Today that's fallen out of favor with the AI guru's because they require knowledge-based programming. Most of the AI research today is with systems that have very little knowledge-based programming. That's vogue are data based systems with zero understanding of the actual training subjects.
 

Thread Starter

rt694157

Joined Dec 15, 2019
57
Can AI or machine learning be use to supply nutrients to plants ? automatic dosing system to feed nutrients to plant

When the plant's are very small they absorb very less nutrient but as they grow they absorb more nutrients. Can correct quantity of nutrient be given to plant every day by AI
 

bug13

Joined Feb 13, 2012
1,863
I have the same question for myself a few years ago. But I couldn't find anything useful in Embedded environment. I was looking at neural network at the time.

But I am pretty impressed by this company, the use camera to do license plate recognition.
https://www.hikvision.com/au-en/
 

nsaspook

Joined Aug 27, 2009
7,494
I have the same question for myself a few years ago. But I couldn't find anything useful in Embedded environment. I was looking at neural network at the time.

But I am pretty impressed by this company, the use camera to do license plate recognition.
https://www.hikvision.com/au-en/
Deep Learning (convolutional neural networks) is both impressive and brittle.

https://arxiv.org/pdf/1911.00927.pdf
Spot Evasion Attacks: Adversarial Examples for License Plate Recognition Systems with Convolutional Neural Networks

https://impillar.github.io/files/inscrypt2019rolma.pdf
We aim at constructing a practical adversarial attack against DL-LPR. The adversarial license plates are expected to be misclassified by DL-LPR but recognized correctly by humans.

https://sites.google.com/view/rolma-adversarial-attack
 
Last edited:

Thread Starter

rt694157

Joined Dec 15, 2019
57
Deep Learning (convolutional neural networks) is both impressive and brittle.
There are many types of dogs and cats in the world If we want to develop a system that identifies a cats or dogs
one camera will require to take an image
What will be required for this system PC/Microcontroller?
What will be required AI/Machine learning?
 

nsaspook

Joined Aug 27, 2009
7,494
There are many types of dogs and cats in the world If we want to develop a system that identifies a cats or dogs
one camera will require to take an image
What will be required for this system PC/Microcontroller?
What will be required AI/Machine learning?
The required datasets for even simple tasks will be large.

https://machinelearningmastery.com/how-to-develop-a-convolutional-neural-network-to-classify-photos-of-dogs-and-cats/
Dogs vs. Cats Dataset Preparation
The dataset can be downloaded for free from the Kaggle website, although I believe you must have a Kaggle account.

If you do not have a Kaggle account, sign-up first.

Download the dataset by visiting the Dogs vs. Cats Data page and click the “Download All” button.

This will download the 850-megabyte file “dogs-vs-cats.zip” to your workstation.
 

andrewmm

Joined Feb 25, 2011
538
There are many many types of AI systems, not least of which the term AI is used for different things over the decades.
One could argue that AI is something that learns form its environment, in which case a PID control is AI !!!

Also by that definition, a system used to identify cats v dogs that does not learn, would not be AI, yet most none engineers would say that is AI.

So to answer your original question, yep I have done AI, ;;;;;

Re your question about using an AI to supply nutrients to plants.
the thing I see is, an AI system needs training or knowledge.
To train , you are going to have to allow plants to die, so the AI learns.
To use knowledge, that "what level of nutrients" is needed.
Which if you know, just a simple dossing system to keep that level would be easiest.

Both systems need some sort of measurement, one needs a "how well the growth is going" the other needs a nutrient meter.

Both could be difficult
 
Top