AI and Machine learning workshop

1st January, 2021

Updated: 28th October, 2022

January 2021

https://teachablemachine.withgoogle.com/train/image

To save some time ahead of tomorrows session ✨

If your providing your own images of your pet/houseplant 🐶🐱🐹🐠🌱

Keep a few images separate from the main folder that we’ll be working with - we will need them to test with! Can you download one of the folders from here - if your pet is a dog don’t download the dog folder etc

If your not providing your own images :cd: Can you download the ‘provided-images’ folder from here.

Training your model

We're going to train our own model to recognise your pet or houseplant!

if your using your own images you'll need to take out a few images

On the left hand side upload your training images. Dont forget to name them

Click “Train”

Upload any of the images from the test folder to see how accurately it has learnt.

(change the dropdown to file)

Click export model -> and click 'upload my model' Copy the sharable link - we'll use this later

Epochs - an epoch is the number of times your training sample has been through the model - 50 epochs is 50 times through the model. Increasing this can make it more accurate but it can lead to more errors

Batch size - is the number of samples used in an iteration of training

Learning rate - a value too small may result in a long training process that could get stuck, whereas a value too large may result in learning a sub-optimal set of weights too fast or an unstable training process.

Accuracy per class - how accurately it predicted the test samples

Confusion matrix - A confusion matrix summarizes how accurate your model's predictions are. You can use this matrix to figure out which classes the model gets confused about.

Accuracy per epoch - the percentage of classifications that a model gets right during training

Loss per epoch - Loss is a measure for evaluating how well a model has learned to predict the right classifications for a given set of samples. If the model's predictions are perfect, the loss is zero; otherwise, the loss is greater than zero

Apply your model

We need to give the page a title!

We also need a place for someone to insert an image.

Once they have uploaded an image we want to display it on the page First of all we need a place to hold the image Secondly we need to do something with the file when its uploaded

Now we need to submit the button First add a button Second add a function to run when you click that button

When we click submit we want to show a loading icon so we realise something is happening

Now we need to load your model - to do this replace the url with your own one from the previous exercise - or just leave it as is to check for guinea pigs!

We now need to fetch the image - theres some jiggery pokery happening here to get it in the correct format for the library we're using

After the result we will need to show the results and hide the loader

We also need to display the results on the index.html page

The prediction function goes away and comes back with the result and updates the results section with the correct information

20 min theory + intro 40 min exercises

give people something they can take away and refer back too

homework

https://www.tensorflow.org/js/demos

2 part

naive model + train diff pretrained model and run it on some new data