In this project, we’ll take pictures of trees, detect the fruits in that tree and decide whether those fruits and vegetables are ripe or not. The purpose is that we’ll install a webcam system that will periodically take pictures and send us a notification when the fruits are ripe.

Goal of the project

The project will help two kind of person. First, those who have vegetables garden at home, second the farmer on improving their process.

learning process

The task can be divided into steps. First, given an image of a tree, we need to detect where the fruits or vegetables are. Second, given the picture of a vegetable or a fruit, we need to decide whether it is ripe or not. The first task is a supervised image segmentation task. Then, the image classification.

First step

As training data, we’ll need pictures where the position of the fruits is annotated.

Second step

As training data, we’ll need several pictures of fruits at different stages of the evolution. Each of these pictures will be labelled ripe or not ripe.

With the data provided by the pictures, we can also add the size of the fruits and their colour as features that can help the algorithm learn more easily. In general, a convolutional neural network is used for tasks involving images.

Going further

Implement the project so that we don't have to look at the fruit until we receive a notification. A notification would be either that the fruit is not growing as expected or that it is ripe and therefore ready to be picked. This could optimize our time if we have a vegetable garden and even more for a farmer.