Chitti app logo

Chitti is better in the app

Coins, streaks and builds — free on Android & iOS

Get the app
Chitti logoChitti

Fruit

7 Minutes · Module 4 - Image Processing

Machine learning fruit classification for students: learn how to train an AI model on Teachable Machine to detect good and bad fruits. A Tamil technology lesson.

  • Uploading good and bad fruit image samples into Teachable Machine
  • Training the image model using default settings like epochs
  • Testing with downloaded images and webcam, then saving the project
What this lesson teachesRead the lesson as text

A machine learning fruit classification model is an artificial intelligence program trained on image samples to recognize the difference between good fruits and bad or spoiled fruits. In packing facilities and supermarkets, this kind of model helps ensure that fresh fruits are sorted into packages while spoiled fruits are set aside for the dustbin.

Setting up and uploading image samples

  • Open the Teachable Machine website, click get started, and select an image project with a standard image model.
  • Rename the first category to good fruit samples and upload images using either direct camera capture or a zip file.
  • Rename the second category to bad fruit samples and upload a collection of pictures showing spoiled fruit.
  • Review the uploaded files to make sure the model has enough visual examples, such as a batch of sixty-four sample images, to learn from.

Training the image model

Before training starts, advanced settings can be checked inside the tool. The platform sets default values of fifty epochs, a batch size of sixteen, and a learning rate of 0.001. Leaving these default settings in place and clicking the train model button begins the training process. When the computer finishes learning from all the uploaded examples, the status updates to confirm that the model is trained.

Testing with files and webcam

  • The system provides options to test the newly trained model either by uploading separate files or by granting camera permissions.
  • Uploading a newly downloaded picture of a good fruit tests the classifier, showing a good fruit result of one hundred percent.
  • Testing another downloaded picture of a damaged fruit accurately classifies it as a bad fruit with ninety-seven percent confidence.
  • Switching the input to a webcam allows live testing by holding a fruit in front of the lens, showing confidence scores between ninety-nine and one hundred percent.

Saving the project

The completed work can be saved by opening the menu and choosing download project as file. This saves the project with a dot tm file extension and keeps all uploaded image samples intact so they can be opened and used again later. The interface also includes an export model option that prepares the finished machine learning model for use in other applications.

Part of the course

A.I. with Teachable Machine

View course →

More in Module 4 - Image Processing

  1. 1FruitNow playing
  2. 2Mask ON & OFF4 Minutes