INTERNATIONAL JOURNAL OF PROGRESSIVE RESEARCH IN SCIENCE AND ENGINEERING, VOL.4, NO.05, MAY 2023.

Fruit Freshness Detection Using Raspberry Pi

Rohit Piske 1, Aparna Nanaware 2, Vishakha Sawant 2, Komal More 2
1Assistant Professor, Department of Electronics & Telecommunication Engineering, VPKBIET, Baramati, Pune, India.
2Student, Department of Electronics & Telecommunication Engineering, VPKBIET, Baramati, Pune, India.
Corresponding Author: rohit.piske@vpkbiet.org

Abstract: - The quality of fruits or vegetables performs an essential role in customer consumption. This project explains the
importance of system for detecting the degradation of fruits that may be required at industrial level, for packaging, etc. This proposed
paper sheds light on the system which is a low cost yet effective fruit quality detecting system in terms of freshness and rottenness
using Raspberry Pi. The developed method can recognize the fresh fruits from rotten fruits and is an advantage over the current
traditional method of identifying the fruits quality which is more time consuming and has reasonably less rate for classifying fruits.
Besides, hand sorting of rotten fruits from fresh fruits is more time consuming, tiring and tedious work. This project focuses on
creating a smart AI camera using Raspberry Pi for effective sorting of fresh fruits from rotten fruits. Fruit quality detection system
involves steps like image capturing, image processing, feature extraction, training, image classification and then recognition. Here,
CNN algorithm (Convolutional Neural Network) is used to train the Machine learning model. A Raspberry Pi camera is used to
capture an image of particular fruits then the trained model detects the quality of fruits and sorts them accordingly as fresh fruit or
rotten fruit. In this project, software which includes Python code is used.

Key Words: Raspberry Pi, AI, Machine Learning, CNN, Python.

I. INTRODUCTION Machine Learning and Deep Learning machine are important

part of technology, which gives several methods to solve any
Fresh fruits are always essential for living a healthful life. Often concerns with less labour cost by training machines. The trained
fruits are rotten and need to be sorted for finding good quality machines in both technologies can smartly work through
amongst them. The traditional way of identifying freshness is provided functions. Identifying fruit with the help of algorithms
by checking every fruit, which is already in use in markets, but or trained machines can be a way to identify fruit quality easily.
it is time consuming and at many times this process results in Hand sorting rotten fruits from fresh fruits in mixed fruits is a
less accuracy. Technology is getting updated day by day. tiring and tedious work. Despite being time consuming, most
Multiple approaches are coming by the researchers to solve and food processing factories carry out this way for making jams
overcome various issues. and other fruit-based products. Hence, to provide a much less
demanding solution, this project focuses on creating a smart AI
camera using a Raspberry Pi Development Board for fresh and
rotten fruit detection.
Manuscript revised May 11, 2023; accepted
May 12, 2023. Date of publication May 14,
2023. II. OBJECTIVE
This paper available online at www.ijprse.com
The main objective is to build a cost-effective system by using
ISSN (Online): 2582-7898; SJIF: 5.59
some of the electronic devices such as Raspberry Pi and
Raspberry Pi camera to give more accurate results then the
present systems are available. This system will also reduce the
hand work for a human being and can give more productivity.

ROHIT PISKE., et.al.: FRUIT FRESHNESS DETECTION USING RASPBERRY PI 70

 INTERNATIONAL JOURNAL OF PROGRESSIVE RESEARCH IN SCIENCE AND ENGINEERING, VOL.4, NO.05, MAY 2023.

III. METHODOLOGY 80% data of our downloaded data sets. Model test to check if
model predicts correctly using remaining 20% of data, and the
3.1 Convolution Neural Network (CNN) final prediction.
Convolutional Neural Networks (CNNs) are a type of deep
learning algorithm commonly used for image processing tasks, IV. RESULTS AND DISCUSSION
including fruit freshness detection. CNNs are specifically
designed to automatically learn patterns and features from Fig 1. Shows the model build for detection and classification of
images, making them well-suited for tasks that require image fruit freshness using image processing.
analysis and classification. The architecture of a typical CNN
consists of multiple layers, including convolutional layers,
pooling layers, and fully connected layers. Here's a brief
overview of each type of layer Convolutional layers: These
layers apply convolution operations to input images, which
involves sliding small filters or kernels across the image to
extract local features. These features may include edges,
textures, or other patterns that are relevant for disease detection
in fruit images. Convolutional layers help the CNN to learn
hierarchical representations of the input images. Pooling layers:
These layers down sample the feature maps generated by the Fig.1. Detection and classification of fruit freshness using image
convolutional layers, reducing their spatial dimensions. processing
Common pooling techniques include max pooling and average Fig.2. Shows the method of handling the fruit to find out
pooling, which help to reduce the computational cost and whether it is fresh or rotten.
increase the model's translation invariance, i.e., the ability to
recognize features irrespective of their position in the image.
Fully connected layers: These layers are traditional neural
network layers, where all neurons are connected to every
neuron in the previous and subsequent layers. Fully connected
layers are typically used in the final layers of a CNN to make
predictions and classify the input images into different disease
classes based on the learned features. The general workflow of
a CNN for fruit freshness detection involves feeding the pre-
processed fruit images as input to the CNN, passing them
through the convolutional, pooling, and fully connected layers
to learn relevant features, and finally making predictions about
the presence or absence of diseases. The CNN is trained using Fig.2. Capturing the fruit
a labelled dataset of fruit images, and the model parameters are Fig.3. Shows the final result of fruit that is rotten banana.
optimized through backpropagation and gradient descent to
minimize the prediction error.

3.2 Dataset Discussion

This project basically classifies the fruit into fresh and rotten.
The data set has been downloaded from Kaggle website. The
data set consists of 2698 sample images. Up next 80-20 split is
performed on the data sets. That is 80% of data set is used for
training Machine Learning model and 20% is used for testing
purpose. Steps performed to build the Machine learning model
are: import predefined libraries, model definition, weights
initialization, compile the developed model, train model using Fig.3. Result of captured fruit

ROHIT PISKE., ET.AL.: FRUIT FRESHNESS DETECTION USING RASPBERRY PI 71

 INTERNATIONAL JOURNAL OF PROGRESSIVE RESEARCH IN SCIENCE AND ENGINEERING, VOL.4, NO.05, MAY 2023.

V. CONCLUSION

Many techniques have been developed over the years for the
fruits-detection and quality maintenance. With the availability
of the machine learning frame works such as Tensor Flow Lite,
the fruit detection has been implemented using Raspberry Pi to
make easy enough for enthusiasts to use machine learning. The
objective of this research is to introduce a system for detecting
freshness of fruits. To provide a much less demanding solution,
this project focuses on creating a 4 smart AI camera using a
Raspberry Pi Development Board for Fresh and Rotten Fruit
Detection and automatically sort it out.
This project will recognize and sort the fruits into fresh and
rotten. This project can overcome the traditional way of
identifying and sorting fruits thereby reducing the manual work
and time. This work can present a system that classifies three
types of fruits named apple, banana, and orange. It will also
separate the fresh and rotten ones. The mechanism that will be
used in this project to classify the fruits is CNN algorithm. This
project will be based on Machine Learning and Deep Learning
algorithms, which can provide real-time information of fruits;
it lowers labor costs. Factories can use it for fruit counting
purposes.

REFERENCES
[1]. S. M. Metev and V. P Shobana G, Reethu, Sudheksha S,
Vinothini K – “Fruit Freshness Detecting System Using
Deep Learning and Raspberry PI”, IEEE, 2022, Issue 15 July
2022.
[2]. Karthickeyan P., Nikesh V., Sanjay V., Dr. K. Devi, M.E.,
Ph.D., Assistant Professor, SRM Valliammai Engineering
College, Kattankulathur-603 2032022, “IOT Based Food
Freshness Detection Using Deep Learning Techniques”,
IRJEDT, 04 Issue: 05 May 2022.
[3]. Miss. P. R. Chavhan, Dr. S.V. Rode - “Colour based Quality
Analysis of Fruits for Automatic Gvrading using Raspberry
PI”, IJIREEICE, ISO 3297, Vol. 6, Issue 3, March 2018.
[4]. Saleem Ulla Shariff, M G Guru Basavanna – “Fruit
Categorization and Disease Detection Using MI Raspberry
Pi Based Fruit Categorization and Quality Maintenance with
Disease Detection Using Ai and Machine Learning”, IJSTR,
VOLUME 9, ISSUE 11, NOVEMBER 2020.
[5]. Prof. Shashant Jaykar, Ritika Umap, Shruti Shende, Kalyani
Hood - “Fruit Degradation Detection System using CMOS
Color Sensor”, IJARSCT, Volume 6, Issue 1, June 2021.

ROHIT PISKE., ET.AL.: FRUIT FRESHNESS DETECTION USING RASPBERRY PI 72