Using Teachable Machine for Potato Leaf Disease Detection

As part of the NAARM MOOC course “AI for Agriculture”, a project needed to be done. The project that I have chosen to do is to use Google’s Teachable Machine for detecting Potato Leaf Disease.

First, I have downloaded the zip file given by the course organizers and have extracted the images to my local.

Then I went to the Teachable Machine site provided in the course resources.

I clicked on the Get Started button and created a new Image Project using the Standard Image Model.

Then I created three classes based on the image data I had: “Healthy”, “Early Blight” & “Late Blight” and uploaded 10 images for each class from my dataset. Then I selected the Train Model button to train the model on my images.

The model trained is not quite accurate as it is labelling any input as “late blight”

So, first I corrected the misspelled name of the “Late Blight” label and added some more images to the “Healthy” and “Early Blight” labels. I also removed some images from the “Late Blight” label. Now, the model is giving correct output for a variety of inputs.

The finished model can be found at the given URL: https://teachablemachine.withgoogle.com/models/XPMlXDuIF/