Noobs Guide to Create Image Classification ML Model using lobe.ai

Hi, In this tutorial I will guide you in creating your First Image Classification Model using Lobe.ai product of Microsoft.

Photo by Jason Leung on Unsplash

With Lobe, anyone can create machine learning models fast and easy — no coding required.

For this tutorial we will be using Fruit Recognition dataset from Kaggle.

Setting up the Environment:

  1. Visit lobe.ai and Download the software on our Windows or Mac.
  2. Download the required Dataset from Kaggle.

How to Train Model:

Open Lobe and Click on New Project, on the Left side there is Label, Train and Play.

To train our model we have to define some data with Labels in our case images with a label.

To add Labels click on Import here are there are 3 ways to create our Label.

  1. Images: In this we have to manually select each images and add a label on it.
  2. Camera: In this we can use are camera to capture and label images.
  3. Dataset: It will accept a Folder which is well structured as per the labels and images.
Dataset File Structure

If you take a look at the kaggle dataset we can see there are 2 folders test and train further going in train you can see your images are well structure which means we can use the 3rd option that is Import Dataset.

So, select the 3rd option and select train folder.

Now, you will get options Label Using Folder Name or Label Manually.

As your selected folder is well structure and label name is folder name itself we will use Label Using Folder Name.

Select the suitable option and click on Import.

As this point of time your dataset will be Importing it can take a long time depending on the size of Dataset and your Processing Power.

Importing Data

Once Importing is Completed it will automatically start Training the Model (training the model can take a lot of time depending on your Dataset Size)

Note:Try to train as much data as you can to improve the accuracy.

In the bottom left you can see 5% of your images are predicted correctly. 95% incorrectly. It has just started training the model but for the training to completed it will improve.

Training Complete

This is the final result after training which is 100% with trained data now we will test it using our test data.

Test Trained Model:

To Test or Play (as per Lobe) with the model you can use your camera or choose image from Device.

Lets drag and drop an image to check if it works.

Hooray 🎉🎉

You ML model is working.

Have you notices Right and Wrong Symbol above image!

We can use this to improve your Model. Every time we test or play with our ML model it will as if it correct or not if not then it will again train the model as per your test images.

This is the reason for which i love Lobe over Teachable Machine.

How to Export:

To export click on 3 line on top left corner and select Export.

Lobe allows you to export in 3 formats TensorFlow, TensorFlow Lite and Local API.

Final Notes

If you have any doubts, feel free to leave a message and don’t hesitate! Don’t be afraid to share this! Thanks!

Flutter & Backend Developer, Junior at Parul University

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store