DeepFruit Fruit Image Classification Dataset

DeepFruit is a fruit image classification dataset jointly released by Prince Mohammed bin Fahd University and other research institutions. The dataset contains 21,122 fruit images of 20 different fruits based on 8 different fruit set combinations. The number of images in each category varies by category. Each fruit appears in 4 or 5 different combinations, and the images are collected from plates of different sizes, shapes, and colors, with different shooting angles, brightness levels, and distances. Example images of all fruit combinations are also attached. The training and test CSV files contain the labels of each corresponding fruit class in each image based on the image file name.

The researchers used the fruit images taken with a Samsung Galaxy S10 smartphone and used preprocessing techniques to rotate, crop, and normalize the images to ensure image consistency and clarity. The dataset was randomly divided into a training set of 80% and a test set of 20% to facilitate model training and evaluation. DeepFruit can be used for research in the fields of fruit detection, recognition, and classification, as well as other innovative applications such as calorie estimation.