Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd
Upload
161 views

Report - Nutrition Analysis Using Image Classification

The document describes an internship report that approves three students - Preethika S, Santhini B, and Saranya E - for successfully completing their project on nutrition analysis using image classification. The project used a convolutional neural network model to predict the type of fruit in an image and its corresponding nutrients. The model was trained on a dataset of 600 images across four fruit categories and tested on 200 new images. The report commends the students for their work and approves their project document and findings.

Uploaded by

poi poi
Copyright
© © All Rights Reserved
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
161 views

Report - Nutrition Analysis Using Image Classification

The document describes an internship report that approves three students - Preethika S, Santhini B, and Saranya E - for successfully completing their project on nutrition analysis using image classification. The project used a convolutional neural network model to predict the type of fruit in an image and its corresponding nutrients. The model was trained on a dataset of 600 images across four fruit categories and tested on 200 new images. The report commends the students for their work and approves their project document and findings.

Uploaded by

poi poi
Copyright
© © All Rights Reserved
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 10

INTERNSHIP REPORT APPROVAL FORM

July 1, 2019
With immense pleasure, this is to approve that the students of Sona college of
technology

Preethika S
Santhini B
Saranya E
successfully completed their Project and Project Report on “NUTRITION
ANALYSIS USING IMAGE CLASSIFICATION” under our guidance.
We are highly impressed with the work that they have done and commend them on
their quick grasping skills. They have shown good intent to learn and have put the
knowledge gained into application in the from of this project. We appreciate the hard
work and commitment shown by them.
We, hereby approve that this document is completely checked and accepted by
SmartBridge Technical Team. Its been an absolute pleasure to educate and mentor
these students. We hope that this document will also serve as a Letter of
Recommendation, to whomsover applied.

We wish them success in all future endeavors and a great career ahead.

GD Abhishek
AI Developer
NUTRITION ANALYSIS USING
IMAGE CLASSIFICATION

Team Name :MIND BENDERS


1. Preethika.S
2. Santhini.B
3. Saranya.E
1. NUTRITION ANALYSIS USING IMAGE
CLASSIFICATION

1.1 INTRODUCTION

Python :Python is anobject-oriented, high-level


programming language and simple, easy to learn syntax readability
and hence it reduces the cost of program maintenance. Pythonalso
supports modules and packagesand also we can use Python for
developing desktopGUIapplications, websites and web applications.
Artificial Intelligence (AI) :Any technique in this world which
mimic the human behavior which has the ability to do intelligence is
called Artificial Intelligence. AI system requires
 Relevant data for training purpose
 Computational power to process the data
AI is an area of computer science that emphasizes the creation of
intelligent machines that work and react like humans. Some of the
activities computers with artificial intelligence are designed for
include. There are several types of Artificial intelligence: ANI-
Artificial NarrowIntelligence ,AGI- Artificial General Intelligence
also known as human-level AIand ASI- Artificial Super Intelligence.
Machine Learning(ML) is a subset of AI techniques. It is the
statistical methods to make machine intelligence and machine
improvement.The various Learning types in Machine learning are as
follows:
 SupervisedMachine Learning
 Unsupervised Machine Learning
 Reinforcement learning.
Deep Learning is a subset of Machine Learning(ML) which make the
computation of multi layer neural network feasible.

1.2 Objectives of Research

The objective of this research project is to find nutritions of the


detected fruit and the name of the fruit using image classification.The
purpose of the project is to classify an object with higher degree of
accuracy by fine tuning the parameters of the network..The main
motto is to reduce the training time and compute complexity of the
network by adding a sub layer after each convolution layer.

1.3 Problem Statement

The problem statement of this project is to find is to predict the kind


of the fruit and and its nutritions.
For example: Predicted fruit : apple
Nutrients :
calories-52
Water-86%

Protein- 0.3 grams

Carbs-13.8 grams

Sugar- 10.4 grams

Fiber- 2.4 grams

The first step of approaching the demanding issue of image


classification is by looking at the available data and preprocessing the
huge amount of data. By preprocessing and training the huge amount
of data we are able to find the prediction of the particular image and
the nutrients of that particular fruit.Hence the user can get to know
the nutrients of the each and every fruit and maintain the healthy
body in order to lead a healthy life.

2.Review of literature

In this project ,CNN algorithm helps to extract the


important features from the images and train according to them.

These neural networks have proven to be successful in many different


real-life case studies and applications like

 Image classification, object detection, segmentation, face


recognition.

 Self driving cars that leverage CNN based vision systems.


 Classification of crystal structure using a convolutional neural
network.

This model involves 4 steps:


 Convolution
 Max pooling
 Flattening
 Full connection

And finally import the image using these packages to find the
predictions:
from keras.models import load_model
from keras.preprocessing import image

The extension for saving the model is .h5 file

3.Data Collection
1.Download the dataset
2.Load the dataset
In this project ,we have used 4 datasets namely
 Apple
 Banana
 Kiwi
 Strawberry
In training dataset ,we have totally 600 images
and in testing dataset we have 200 images.

4.Methodology

4.1 Exploratory Data Analysis


Figure 4.1.1
4.2 Data Modelling

Convolution Neural Network (CNN) algorithm is applied to this


model .This helps us to extract the important features from the images
and train them.

It has four steps namely


1. Convolution
2. Max pooling
3. Flattening
4. Full connection

In convolution operation,the input image is convoid with feature


detector or filter to get a feature map.By applying convolution
operation,the size of image is reduced we may lose some information
but feature detector or filter will helps us to extract main features
from image and remove unwanted features.
Max pooling is a technique which helps us to avoid overfitting of
data and helps us to avoid special ingredients and distortions in data.
Flattening layer converts multi dimension pooled feature map to
single dimension pooled feature map.Flattening layer is the input
layer.

Start initializing the parameters and finally train the model with Keras
fit() function.The model trains for 20 epochs.
Testing is similar to training, except that we don’t need to compute
gradients and training targets. Instead, we take the predictions from
network output, and combine them to get the real detection output.
Fit_generator is used to find the training and validation accuracy.
Image data generator is a class in keras.preprocessing package to
apply some image processing to the images.

In this project the accuracy error is also good,so that the model is
perfect.

5. References

1) https://www.geeksforgeeks.org/image-classifier-using-cnn/

2) Jupyter Notebook – Conv_Net

3) Wikipedia – Convolutional Neural Networks

4) Stanford Course – cs231n

5)
https://pdfs.semanticscholar.org/f8dd/7286ed56e8d9e5144158f0b79a
c5559c048b.pdf
6)
https://cv-tricks.com/object-detection/faster-r-cnn-yolo-ssd/

6. Conclusion

Convolutional neural networks (CNNs) have accomplished


astonishing achievements across a variety of domains, including
medical research, and an increasing interest has emerged in
radiology.Detection is harder than classification, since we want not
only class probabilities, but also localizations of different objects
including potential small objects. Using sliding window together with
a good classifier might be an option, however, we have shown that
with a properly designed convolutional neural network, we can do
single shot detection which is blazing fast and accurate. Hence using
Convolution Neural Network(CNN)algorithm we have predicted the
nutrition of the fruit which is very helpful for the user to maintain
their healthy lifestyle.

You might also like