Create simple Image Classification model / Pose detector model with Google Teachable Machine

#machinelearning #imageclassification #posedetection #ai

We are going to create a simple pose detector (thumbs up, victory and thumbs down) model with the help of google's teachable machine leaning website.

Open a web browser and goto this website : https://teachablemachine.withgoogle.com

You should see something like this webpage. Click on get started.

Select image project:

Choose standard image model:

You should see some sections including 2 class section a training section and a preview section.

Here the classes are the different types we want to recognise. In our case thumbs up, victory and thumbs down. You can change the names of classes according to your liking. I named the first class as "ThumbsUp".

Now, you should see expansion of ThumbsUp section similar to this:

Click on Hold to record to give image input to the model. There should be a pop up about accessing the webcam of your device. You should allow this to take your poses.

On holding the button, it will keep taking photos of your poses. When you are confident about the number in my case 300+ just let the button free.

These are the poses of ThumbsUp class.

Now you can do the same thing for the Victory pose, First rename if you want then take poses close to the first pose.

To create a new pose click add class button and rename the class
then do same as previous.

Now, We have our classes. Goto the training portion.
You can start training with the default value. If you want to change the default values. You can click the advance section. Here you can change several parameters. After you satisfied with the numbers. Click on training button. It will take some time. Keep patience.