Custom Vision Machine Learning Example – Baseball themes


everyone professor Yates here with
another example of a machine learning algorithm in model in this particular
example that I’m going to show you I train the Microsoft Azure custom vision
AI tool to recognize baseball so baseball’s baseball bats balls bats
gloves hats and jerseys as well as players so what I’m going to try to show
you here is that after training the Microsoft Azure custom vision model I
can then test it by writing that I wrote a small application to do this test the
machine learning model out to see if it can recognize images that it hasn’t seen
before so what I’m going to do is based on this training that I’ve done with
these images up here I’m gonna actually send it images from baseball that are
that it has never seen before so the idea here is again it’s a supervised
training model supervised machine learning where we have trained it on
particular kinds of images and then we’re going to send it images so it’s a
cloud-based application on Microsoft Azure so the models were built on Azure
in the cloud and now I’m going to test it using a quick web application that I
built what I’m going to do is I’m going to test it on images it hasn’t seen
before and I’m gonna test it on images that it shouldn’t recognize this
baseball for instance basketball alright so I’m gonna send it some basketball
pictures and and see what it comes back with so let’s take a quick look at this
application so here is the web application and you can see the the
baseball motif there and I’ve actually got a player up as you see right here on
the screen and basically what I’m gonna do is I’m going to open it up so I’m
gonna say ok custom learning machine learning image what is this image right
here so based based on it being trained in baseball it should take a look at
that image and you know what that is not baseball so
let’s take a look at see what it says so what does in this model came back with
is is the model predicted this image is negative meaning this is not anything
that I know of nothing that I’ve been trained on so chance essentially what
we’ve done here again is train the model on a series of images that we wanted to
be able to recognize going forward in the future and then anything other than
that I’ve programmed to say you know what it is in baseball so that’s when we
get this negative prediction right here so essentially what this is doing is a
prediction there’s a 100% predictive the models a hundred percent sure that this
is not baseball all right let’s try something else so I’m gonna choose a
file I’m gonna choose a player so I’ve also trained it again on players in my
machine learning model let’s let’s ask it to see an image that it hasn’t seen
before and see what it says it is so let’s open this image right here and
we’re gonna ask the model what is it well this is predicted to be a hundred
percent of player all right so so far my model is doing pretty well and again
this is all running on the Microsoft Azure cloud all right so let’s choose
another one let’s try this hamburger right here the köppen on that and what
is this and again 100% negative so what we’re seeing here is another application
of machine learning Microsoft Azure and all the cloud platforms of a pretty
robust capability in terms of being able to Train it in this particular use case
I’m training it on the in terms of baseball but we could literally train
the model the image model to visualize any kind of image based on training and
then test the model later on to see if it actually works again this is another
example of supervised learning so with that continue on with the module and
hopefully you’ve enjoyed this little segment

Leave a Reply

Your email address will not be published. Required fields are marked *