For our final project of the Artificial Intelligence (AI) course, we've created an application called Frezz, which is an application that can sort fruits through an artificial intelligence approach. This application is motivated by the process of quality control of fruit freshness which is generally done manually during the process of supplying fruit (farmers to supermarkets) and controlling sales (supermarkets to customers). In this app, we focused on apples, bananas and oranges; creating 6 categories: fresh apples, rotten apples, fresh bananas, rotten bananas, fresh oranges and rotten oranges. For the fruit dataset, we use the dataset from Kaggle. Then, the dataset is used for model training in Google Colab by integrating Tensorflow and Keras. The model that has been trained obtains an accuracy level with a value of 0.984000027179718. In developing the web application, we used React and Tensorflow.js to integrate the model into the Frezz web application. Therefore, the website is able to display the results of fruit freshness detection where the input (photo/camera) can be categorized into one of the 6 categories in the dataset.
Solution
Frezz uses deep learning techniques to analyze the images of the fruits and determine their freshness level. The project includes a dataset of images of fresh and non-fresh fruits, which was used to train a convolutional neural network (CNN) model. The trained model is then used to predict the freshness of fruits from input images.
Dataset
The dataset used to train the CNN model in Frezz was sourced from Kaggle. The dataset includes images of fresh and non-fresh fruits, specifically apples, oranges, and bananas.
Contributors
This project was completed as a group project for the Artificial Intelligence course. The contributors are:
- Alyza Pramudya
- Shafa Amira
- Dierta Pasific
- Jacqueline Abyasa