Inspired by the Google Teachable Machine - Image Project, this is an alternative open-source version with the same features.
You must grant camera permission to use the application.
The Teachable Machine is a machine learning application based on ml5js that runs in a web browser. It classifies objects using training from provided images. Users can supply training images directly through their computer's built-in camera.
To make the Teachable Machine functional, you need at least two classes. Click the "+" button to add a new class.
Clicking the "Pencil" icon next to each class name allows you to change the class name.
In the example below, we will use 3 classes: 1) Blank, 2) Panda, and 3) Husky.
The main video displayed on the left is live-streamed from your default system camera. Use the "Camera" icon next to each class to capture the current image for that class. You can also use the "Trash" button in the top-right corner of each image to remove unwanted images.
It is recommended to add more than 16 images for each class.
After adding training images for all classes, click the "Training" button to start the training process. A progress bar will appear on the bottom left to indicate current progress.
Once the training progress is complete, the Teachable Machine will enter classification mode. The application will continuously detect objects in the images captured from the main video and classify them based on the training data.
The progress bars on the bottom left will display the potential class of the current object along with the confidence level.
