A end to end example of using Deep Learning for the task of... classifying water melons. You knock on a melon, the program records the sound, and predicts if it is sweet and soft or not.
Of course, you want to do it in the store when buying the thing, and it means you can not go there carrying your computer connected to Google Colab. So we will design and train the NN on your home PC, and then port it to an Android phone.
Note that Google keeps changing the rules for software uploaded to Google Play, so the link yo Melonaire in a Google Play is most likely dead. I have neither time not motivation to keep it updated. You can always compile a new version on a latest Android Studio (or use the old one if you don't want to comply to all changes they introduced), and it will work. It is a tutorial, anyway. To learn to classify water melons will take you (a human) just few attempts, so you do not need software for that.
A dogs species classifier I wrote for a Kaggle competition.