Skip to content

srbhr/Fruits_360

Repository files navigation

Fruits_360

Fruits Detection using CNN.

Dataset used :

Fruits 360

A dataset of images containing fruits and vegetables

Dataset properties
  • Total number of images: 82213.
  • Training set size: 61488 images (one fruit or vegetable per image).
  • Test set size: 20622 images (one fruit or vegetable per image).
  • Multi-fruits set size: 103 images (more than one fruit (or fruit class) per image)
  • Number of classes: 120 (fruits and vegetables).
  • Image size: 100x100 pixels.

The Dataset Can be found over : https://www.kaggle.com/moltean/fruits and https://github.com/Horea94/Fruit-Images-Dataset

This is the work of Horea Muresan, Mihai Oltean, Fruit recognition from images using deep learning, Acta Univ. Sapientiae, Informatica Vol. 10, Issue 1, pp. 26-42, 2018.

The paper introduces the dataset and an implementation of a Neural Network trained to recognized the fruits in the dataset.

How to use this

Requirements

The requirements.txt file, has all the packages that were in the environment at the time of training.

  • Tensorflow 2.0 (Tensorflow-GPU was used)
  • Keras 2.3.1
  • Matplotlib
  • Numpy

Usage

The Images to be predicted are put under the fruits/test_images folder.

This model is pretrained with and weights is a H5py file. Named 'Fruits_360.h5'.

The fruits.py file contains the Network Model and was used to train it.

About

Fruits Detection using CNN.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published