CNN and Data Augmentation Based Fruit Classification Model

Identification of different fruits is still an important and tedious task in the food processing industry. Workers not only have to identify the fruit but also have to make decisions to determine if the fruit is edible or not. Classification of fruits into edible and non-edible classes can be proved as a very important aspect in such industry. In our proposed system four fruits are classified namely, Banana, Papaya, Mango, and Guava into three stages raw, ripe, and over-ripe using Convolutional Neural Networks. In the model, a dataset of local fruits is used and studied their life cycle in different stages. In this, an accuracy of 97.74 % in 8 epochs with 0.9833 validation precision was achieved. The same model can be implemented using real-time images captured using cameras to identify the edibility of fruit which can prove to be helpful for everyone.