Fruit Ripeness Detection React Application

This app can be used for real time object detection by leveraging TensorFlow Object Detection API, React, Tensorflow JS and IBM Cloud. This repository contains the code and Google Colab notebook for training a detector of banana ripeness.

Problem Statement

The model is used to detect the ripeness process of bananas. Images used for training and test are taken from Roboflow. There are 5.3K images for training and 757 images for test.

The corresponding labels are:

• Freshripe
• Freshunripe
• Overripe
• Ripe
• Rotten
• Unripe

Model output examples

Live web app detection

• Test image from web app

Lessons learned

• Input image complexity In some of the training images, it's difficult to distinguish different stages of banana ripeness. Data augmentation process needs to be improved to consider more cases and improve the model detection accuracy.