See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/360726233

PLASTIC DETECTION AND CLASSIFICATION USING DEEP LEARNING A

PROJECT REPORT (Project Work I -Phase I)

Conference Paper · May 2022

CITATIONS READS

0 3,314

3 authors:

Sivaram Cp Mukesh Sekar

Bannari Amman Institute of Technology Bannari Amman Institute of Technology
2 PUBLICATIONS   0 CITATIONS    2 PUBLICATIONS   0 CITATIONS   

SEE PROFILE SEE PROFILE

Rajkhishore K S
Bannari Amman Institute of Technology
3 PUBLICATIONS   0 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

plastic waste classification and detection using deep learning View project

All content following this page was uploaded by Rajkhishore K S on 20 May 2022.

The user has requested enhancement of the downloaded file.

 i

PLASTIC DETECTION AND CLASSIFICATION

USING DEEP LEARNING

A PROJECT REPORT
(Project Work I - Phase I)

RAJKHISHORE S
(REG NO:182IT201)

MUKESH S
(REG NO:182IT179)

SIVARAM C
(REG NO:182IT222)

in partial fulfilment of the requirements

for the award of the degree

of

BACHELOR OF TECHNOLOGY
INFORMATION TECHNOLOGY
DEPARTMENT OF INFORMATION TECHNOLOGY
ENGINEERING COLLEGE

(Autonomous)
SATHYAMANGALAM– 638 401

JUNE 2022
 ii

DEPARTMENT OF INFORMATION TECHNOLOGY

BANNARI AMMAN INSTITUTE OF TECHNOLOGY

(Autonomous)
SATHYAMANGALAM – 638401
JUNE 2022

BONAFIDE CERTIFICATE

This is to certify that the Project report entitled PLASTIC DETECTION AND

CLASSIFICATION USING DEEP LEARNING is the bonafide record of project

work done by RAJKHISHORE S (Register no: 182IT201), MUKESH S (Register

no: 182IT179), SIVARAM C (Register no:182IT222) in partial fulfilment of the

requirements for the award of the Degree of Bachelor of technology in information

technology of Anna university Chennai during the year 2021 - 2022.

SUPERVISOR HEAD OF THE DEPARTMENT

(Signature with seal)
Date:
 iii

DEPARTMENT OF INFORMATION TECHNOLOGY

BANNARI AMMAN INSTITUTE OF TECHNOLOGY

(Autonomous)

SATHYAMANGALAM – 638401

JUNE 2022

DECLARATION

We affirm that the Project Report titled PLASTIC DETECTION AND CLASSIFICATION
USING DEEP LEARNING being
submitted in partial fulfilment of the requirements for the award of Bachelor of Engineering
is the original work carried out by us. It has not formed the part of any other project report or
dissertation on the basis of which a degree or award was conferred on an earlier occasion on
this or any other candidate.

Date:
RAJKHISHORE S
(REG.NO:182IT201)

MUKESH S
(REG.NO:182IT179)

SIVARAM C
(REG.NO:182IT222)

I certify that the declaration made by the above candidates is true to the best of my knowledge.

Date: Name and Signature of the Supervisor with seal

 iv

ABSTRACT

Plastic isn’t as simple as everyone assume. The world now produces more than 380
million tons of plastic, half of this is utilized to make one-use items such as polythene bags,
mugs, and straws. Every year, at least 8 million tons of plastic wind up in our oceans, which
end up as pollutants, entering our natural environment. Thus, it is necessary to handle such
kind of massively produced material. There are 7 major categories of plastics. Every single
one of them is distinct from the others. Some are reusable, but others could leach harmful
compounds if exposed to extreme temperatures.

Some materials are easily recyclable, while others require more complex and delicate
processing during the recycling process. Although scientists are actively striving to develop
the optimum method and strategy for recycling all of those forms of plastic, recycling
programs mostly accept Polyethylene Terephthalate (1-PET) and High-Density Polyethylene
(HDPE) (2-HDPE).

Plastic categorization by hand is a time-consuming and costly operation. That’s why

we need an automated sorting technique in order to classify it to increase recycling, image
processing and artificial intelligence, particularly deep learning technique, are being used.
This is an energy conserving technique. Plastic components are the most problematic in-
home waste, and the most common forms are polyethylene, polypropylene, and polystyrene.
We presented a strategy for identifying garbage in portable gadgets that could be useful in
resolving urban waste issues.
 v

ACKNOWLEDGEMENT

We could like to enunciate heartfelt thanks to our esteemed chairman

Dr. S V Balasubramaniam and the respected Director Dr. M P
Vijaykumar, for providing excellent facilities and support during the
course of study in this institute.

We are grateful to Dr. S Daniel Madan Raja , Associate Professor

and Head of Information Technology for his valuable suggestions to
carry out the project work successfully.

We wish to express our sincere thanks to Project Co-ordinator Ms. P

Ponnila, Assistant Professor, Information Technology for her
constructive ideas, inspirations, encouragement and much needed
technical support extended to complete our project work.

We wish to express our sincere thanks to Faculty guide Mrs. B.

Santhiya , Associate professor , Information Technology for his
constructive ideas, inspirations, encouragement, excellent guidance and
much needed technical support extended to complete our project work.
We would like to thank our friends, faculty and non-teaching staff who
have directly and indirectly contributed to the success of this project.
 vi

TABLE OF CONTENTS

CHAPTER NO TITLE PAGE NO

ABSTRACT iii

TABLE OF CONTENTS vi

LIST OF ABBREVIATIONS vii

LIST OF FIGURES ix

1 INTRODUCTION 1

1.1 OVERVIEW OF PLASTIC WASTE 1

1.2 CONVOLUTIONAL NEURAL NETWORKS 2

1.3 FUTURE FORECASTING 3

1.4 DEEP LEARNING 3

1.5 OBJECTIVE 4

2 LITERATURE SURVEY 6

2.1 EXISTING SYSTEM 11

2.2 DRAWBACKS OF EXISTING SYSTEM 12

2.3 PROPOSED SYSTEM 12

2.4 ADVANTAGES OF PROPOSED SYSTEM 13

3 SYSTEM SPECIFICATION 14

3.1 HARDWARE REQUIREMENTS 14

3.2 SOFTWARE REQUIREMENTS 14

4 SOFTWARE DESCRIPTION 15

4.1 FRONT END PYTHON 15

4.2 FEATURES OF PYTHON 15

4.3 GOOGLE COLAB 16
 vii

5 MODULE DESCRIPTION 19

5.1 PROBLEM DEFINITION 19

5.2 OVERVIW OF THE PROJECT 19

5.3 MODULE DESCRIPTION 20

6 RESULTS AND DISCUSSION 24

7 CONCLUSION 25

APPENDIX 26

REFERENCES 28
 viii

LIST OF ABBREVIATIONS

HDPE-high-density-polyethylene
PET-polyethylene-terephthalate
LDPE-low-density-polyethylene

PVC-polyvinyl-chloride
PS-polystyrene
PP-polypropylene
 ix

LIST OF FIGURES

FIGURE No. TITLE PAGE No.

5.1 Dataset of Plastics 27

5.2 Image Before Segmentation 27

5.3 Image After Segmentation 28

5.4 System Flow Diagram 28

6.1 Output of Prediction 29

6.2 Output of Classification 30

 1

CHAPTER 1

INTRODUCTION

1.1 OVERVIEW OF PLASTIC WASTE

Solid wastes are being a main threat in environmental protection. In which plastics are
being a serious threat. Since its production rate is enormously high and reusing of these plastics
are low. Because plastic does not biodegrade or absorb back into the environment, it can linger
in landfills for hundreds of years, breaking down into smaller and smaller particles known as
micro plastics. That's why avoiding single-use plastic and minimising the amount of plastic we
use is the single most effective approach to keep plastic out of landfills and seas. Indonesia is the
second-largest source of plastic pollution in the world, generating roughly 24,500 tons of plastic
waste per day. Irrational material management is the primary cause of severe waste production.
Some plastics are single-time usage plastics which can be avoided but some are more difficult to
avoid from our daily life such as our cell phones and our vehicles and so on. When we can't limit
our plastic consumption, we should think of ways to reuse it.

The 3R Initiative intends to promote the "3Rs" (reduce, reuse, and recycle) internationally
in order to create a sustainable material-cycle society by maximizing resource and material
efficiency. The 3R’s principle mainly focuses on reusing, reducing waste and recycling resources.
Reducing waste means choosing things with care to reduce the waste produced. Reusing waste
involves repeated usage of the items that are produced. Recycling the waste means using it as the
raw material for developing another product. Non-waste technology (NWT) is centred on
avoiding waste and maximising the use of basic materials. This entails a number of technological
processes that lead to total waste management and removal that are not detrimental to the
environment. This technology should not be used to dispose of garbage.

The implementation of this financially justified. By lowering energy-consuming waste

treatment operations, it is possible to reduce the consumption of power, heat, or technology. Its
 2

primary concept is to maximise the reuse of the same materials while lowering the cost of
processing them. In general, items are recycled in two places. One is at the manufacturing level,
while the other is at the post-consumer level. From consumer end we process the material and
use it as the raw material for the second time in production of another product. Because not all of
the materials in each group are recyclable, advanced technologies must be used to determine the
type of material in each group .to recycling process of the plastics. We presented a technique for
trash recognition that can be used in portable devices, sorting the materials eligible for recycling
from municipal solid garbage is a time-consuming and costly procedure. As a result, a technique
free of these flaws should be constructed

1.2 CONVOLUTIONAL NEURAL NETWORKS

CNN is a deep neural network aimed to solve instance segmentation problem in machine
learning or computer vision. In other words, it can separate different objects in a image or a video
and image, it gives you the object bounding boxes, classes and masks.

Convolution, it is an important and essential layer .it is a linear pointwise multiplication

of two functions to generate the third function. The ultimate goal of convolution in CNN is to
extract features from the input image and here 4 convolutional layers with filters in the range of
16,32,64,128 are used. Here, 4 convolutional layers are also implemented.

In Pooling layer, pooling is performed. There are 3 types of pooling. They are avgpooling,
minpooling, maxpooling. In this project, maxpooling is used with a pooling size is 2x2. It is a
function used to reduce the number of parameters and computation in the network. In pooling, we
can down sample the pixels to make image fewer parameters to characterize the image.

Dropout is used. It means ignoring the units. It is done to prevent overfitting problems.
Overfitting problem is when learning details to an extent will have a negative impact on
performance. In this project, drop out is done in range of 0.3 to 0.4. Fully connected layers are
those layers where all the inputs from one layer are connected to every unit of the next layer. Fully
connected layers are used between convolution al layer and output layer.
 3

1.3 FUTURE FORECASTING

In future work, we plan to apply the concepts and best practices of deep learning described
in this survey to other medical fields that have not fully utilized this technology. This survey aims
to encourage many researchers to conduct deep learning experiments and apply the model of deep
learning in the field of computer vision involving medicine, thereby achieving smart and
convenient development for the medical industry.

The development of skin disease diagnostic technology and the process of traditional
medical diagnosis is outlined. The machine learning-based image recognition of skin disease and
investigate the results of deep learning-based research in the field of skin disease recognition is
also discussed. Forty-five relevant papers have been identified to obtain their concerns about the
areas and skin disease types. These papers are utilized as basis to study the used data source, the
pre-processing and data expansion techniques, the technical details of the models, and the
performance indicators’ overall performance. Deep learning models are widely used in skin
disease recognition. Researchers often use the multi model fusion technology to improve the
performance of models. Forecasts are based on opinions, intuition, guesses, as well as on facts,
figures, and other relevant data. All of the factors that go into creating a forecast reflect to some
extent what happened with the business in the past and what is considered likely to occur in the
future.

1.4 DEEP LEARNING

Deep learning is an artificial intelligence function that imitates the workings of the human
brain in processing data and creating patterns for use in decision making. Deep learning is a subset
of machine learning in artificial intelligence that has networks capable of learning unsupervised
from data that is unstructured or unlabelled also known as deep neural learning or deep neural
network. Deep-learning methods are representation-learning methods with multiple levels of
representation, obtained by composing simple but non-linear modules that each transform the
representation at one level starting with the raw input into a representation at a higher, slightly
more abstract level. The key aspect of deep learning is that these layers of features are not designed
by human engineers: they are learned from data using a general-purpose learning procedure.
 4

Deep learning, also known as deep neural networks or neural learning, is a form of
artificial intelligence that seeks to replicate the workings of a human brain. It is a form of machine
learning, with functions that operate in a nonlinear decision-making process. Deep learning occurs
when decisions are made on unstructured data without supervision. Object recognition, speech
recognition, and language translation are some of the tasks performed through deep learning.

Deep learning has evolved hand-in-hand with the digital era, which has brought about an
explosion of data in all forms and from every region of the world. This data, known simply as big
data, is drawn from sources like social media, internet search engines, e-commerce and online
cinemas, among others. This enormous amount of data is readily accessible and can be shared
through fine tech applications like cloud computing. Deep learning is used across all industries for
a number of different tasks. Commercial apps that use image recognition, opensource platforms
with consumer recommendation apps, and medical research tools that explore the possibility of
reusing drugs for new ailments are a few of the examples of deep learning incorporation.

1.5 OBJECTIVE

Every year, at least 8 million tons of plastic wind up in our oceans, which end up as
pollutants, entering our natural environment. Thus, it is necessary to handle such kind of massively
produced material. There are 7 major categories of plastics. Every single one of them is distinct
from the others. Some are reusable, but others could leach harmful compounds if exposed to
extreme temperatures.Some materials are easily recyclable, while others require more complex
and delicate processing during the recycling process. Although scientists are actively striving to
develop the optimum method and strategy for recycling all of those forms of plastic, recycling
programs mostly accept Polyethylene Terephthalate (1-PET) and High-Density Polyethylene
(HDPE) (2-HDPE). This classification of plastics for recycle, reuse or reduce is a necessary thing
to avoid the greenhouse effect and ozone layer depletion, which Is a difficult thing to achieve
through manual power.

The main idea is to reduce the waste and to recycle and promote the reuse of the same
materials while lowering the cost of processing them. Thus, consumer waste can be recycled by
combining the secondary use of raw materials with a change in their condition and composition
The fundamental issue addressed in this article is the development of an automatic technique for
 5

sorting plastic waste. This necessitates sorting waste into fractions other than metal, bio, plastic,
paper, and glass. Because not all of the components in each group are acceptable for reuse today,
advanced procedures must be used to distinguish the type of material in each group This can sort
into the four categories listed above, PS, PP, PE-HD, and PET, and can be used in a sorting factory
or at home by residents. Plastic categorization by hand is a time-consuming and costly operation.
Although the single output of a discriminative regression model is misleading, the output of a
regression model actually relates to a well-known probability distribution, the Gaussian
distribution. As it turns out, the output of a discriminative regression model represents the mean
of a Gaussian distribution (a Gaussian distribution is fully defined by a mean and a standard
deviation). With this information, you can determine the likelihood of each real value given the
input. This study mainly summarizes the research and application progress of Plastic image
recognition based on deep learning, briefly introduces the methods and status of plastic
recognition. It introduces the development history of Plastic detection focuses on the research
progress of plastic image recognition based on deep learning summarizes the full text and discusses
plastic image recognition's future research trends based on deep learning.

Only the mean value of this distribution is typically modelled, and the standard deviation
of the Gaussian is either not modelled or chosen to be constant across all plastic. In discriminative
regression models, θ thus defines a mapping from image recognition to the mean of a Gaussian
from which is sampled in image format of plastic. The mean value is almost always chosen when
making a decision.
 6

CHAPTER 2

LITERATURE REVIEW

“Solid waste issue: Sources, composition, disposal, recycling, and valorization” 2018 By
panelHussein I.Abdel-ShafyaMona S.M.Mansou ,Disposal of solid wastes is a stinging and
widespread problem in both urban and rural areas in many developed and developing countries.
Municipal solid waste (MSW) collection and disposal is one of the major problems of urban
environment in most countries worldwide today. MSW management solutions must be financially
sustainable, technically feasible, socially, legally acceptable and environmentally friendly.

“A review on automated sorting of source-separated municipal solid waste for recycling”

2016 By Sathish Paulraj Gundupalli 1, Subrata Hait 2, Atul Thakur 3 A crucial prerequisite for
recycling forming an integral part of municipal solid waste (MSW) management is sorting of
useful materials from source-separated MSW. Researchers have been exploring automated sorting
techniques to improve the overall efficiency of recycling process. This paper reviews recent
advances in physical processes, sensors, and actuators used as well as control and autonomy
related issues in the area of automated sorting and recycling of source-separated MSW

“Influence of shape and size of the particles on jigging separation of plastics mixture,”
2015 Fernando Pita 1, Ana Castilho 2 Plastics are popular for numerous applications due to their
high versatility and favourable properties such as endurance, lightness and cheapness. Therefore,
the generation of plastic waste is constantly increasing, becoming one of the larger categories in
municipal solid waste. Almost all plastic materials are recyclable, but for the recycling to be
possible it is necessary to separate the different types of plastics. The aim of this research was to
evaluate the performance of the jig separation of bi-component plastic mixtures. For this study six
granulated plastics had been used: Polystyrene (PS), Polymethyl methacrylate (PMMA),
Polyethylene Terephthalate (PET-S, PET-D) and Polyvinyl Chloride (PVC-M, PVC-D). Plastics
mixtures were subjected to jigging in a laboratorial Denver mineral jig. The results showed that
 7

the quality of the jigging separation varies with the mixture, the density differences and with the
size and shape of the particles.

“Intelligent solid waste processing using optical sensor-based sorting technology,” 2010 Jiu
Huang; Thomas Pretz; Zhengfu Bian All Authors Solid wastes are always collected as mixtures of
different materials. They get crushed, classified and sorted in solid waste treatment plants. Among
these processes the sorting is the determining step for recycling and reuse. Traditional sorting
technologies like magnetic sorting and eddy current sorting are only able to process some special
kinds of ingredients of waste mixture roughly, such as the separation of ferrous and non-ferrous
metals. Since there exist corresponding force fields between waste particles and separators. Some
other properties of the solid particles such as the colours, shapes and texture features could also be
considered as sorting criterions but there is no sufficient force field between these properties and
separators. In this paper, an indirect sorting process by using optical sensor and mechanical
separating system was developed and introduced.

“Advanced waste-splitting by sensor-based sorting on the example of the MTPlant

oberlaa,” 2020 Janusz Bobulskia and Mariusz Kubanekb Garbage management is a challenge for
the whole world. The plastic waste might be automatically selected on the sorting lines businesses
for waste disposal by using methods of Computer Vision. Manual sorting of garbage is a tedious
and expensive process, which is why scientists create and study automated sorting techniques to
improve the overall efficiency of the recycling process.

“Upgrading the quality of mixed recycled aggregates from construction and demolition
waste by using near-infrared sorting technology,” 2015 InigoVegas,Kris Broosb,Peter Nielsenb,
Oliver Lambertz,Amaia Lisbonaa Recycled aggregates of high-purity, guaranteeing optimal
technical and environmental performance, are required for high-grade construction applications
such as concrete. The main problem constituents causing a decrease in the quality of recycled
aggregates to be used in high grade applications are: organic material, gypsum and autoclaved
aerated concrete (AAC). This paper studies the potential of Near Infrared (NIR) sorting
technology to improve the quality of mixed recycled aggregates. Tests were carried out by using
samples of mixed recycled aggregates collected in different EU countries (Germany, Sweden,
Spain and Italy).
 8

“Case study: interpretability of fuzzy systems applied to nonlinear modelling and control,”
2017 Krzysztof Cpalka Fuzzy systems have their limitations which result from a number of factors
including a limited ability to assure their interpretability in various practical problems. If
interpretability is not of paramount importance, then we can consider using methods from the
“black box” group (e.g., artificial neural networks with a teacher). However, if interpretability is
of a great importance, then some dedicated approaches and algorithms should be developed.

“Real-time hyperspectral processing for automatic nonferrous material sorting,”

2012Artzai Picon, Aranzazu Bereciartua, Jone Echazarra, Ovidiu Ghita, Paul F. Whelan, Pedro M.
Iriondo The application of hyperspectral sensors in the development of machine vision solutions
has become increasingly popular as the spectral characteristics of the imaged materials are better
modeled in the hyperspectral domain than in the standard trichromatic red, green, blue data. While
there is no doubt that the availability of detailed spectral information is opportune as it opens the
possibility to construct robust image descriptors, it also raises a substantial challenge when this
high-dimensional data is used in the development of real-time machine vision systems.

“Industrial application for inline material sorting using hyperspectral imaging in the NIR
range,” 2005 Petra Tatzer,Markus Wolf,Thomas Panner Spectral imaging is becoming
increasingly interesting not only for agricultural use but also for industrial applications.
Wavelengths in the near infrared (NIR) range, in particular, can be used for materials
classification. However, sorting paper according to quality is a very difficult task due to the close
similarities between the materials. This work describes the development of a unique industrial
inline material sorting system which uses the spectral imaging technique.

“Sorting of polypropylene resins by color in MSW using visible reflectance

spectroscopy,” 2010 S. M. Safavi, H. Masoumi, S. S. Mirian, and M. Tabrizchi, In this paper, an
automated sorter is proposed for distinguishing polypropylene (PP) plastics based on their color.
This sorting system uses visible (VIS) reflectance spectroscopy to separate PP resins according
to their colors. A “Three-Filter” identification algorithm was developed to recognize the PP color
 9

(blue, red, green, white or yellow), and accordingly, give the command for throwing or not
throwing PP to a series of electro pneumatic valves.

“ utilization of hyperspectral imaging for impurities detection in secondary plastics,” 2012

S. Serranti, A. Gargiulo, G. Bonifazi, A. Toldy, S. Patachia, and R. Buican, The systematic
identification of impurities inside secondary plastics flow streams can be considered as one of the
key issues to certify and to classify waste plastics fed to recycling plants and to perform a full
control of the resulting processed fractions, that have to comply with market demands on grade
and purity of the recovered products when compared with virgin streams. HyperSpectral Imaging
(HSI) can represent an optimal, reliable and low costs answer to reach the previous mentioned
goals.

“Characterization of post-consumer polyolefin wastes by hyperspectral imaging for

quality control in recycling processes,” 2011 S. Serranti, A. Gargiulo, and G. Bonifazi, In this
paper new analytical inspection strategies, based on hyperspectral imaging (HSI) in the VIS–NIR
and NIR wavelength ranges (400–1000 and 1000–1700 nm, respectively), have been investigated
and set up in order to define quality control logics that could be applied at industrial plant level
for polyolefins recycling.

“Classification of polyolefins from building and construction waste using NIR

hyperspectral imaging system,” 2012 S. Serranti, A. Gargiulo, and G. Bonifazi, This work was
carried out to develop a hyperspectral imaging system in the near infrared (NIR) range (1000–
1700 nm) to classify polyolefin particles from complex waste streams in order to improve their
recovery, producing high purity polypropylene (PP) and polyethylene (PE) granulates, according
to market requirements. In particular, hyperspectral images were acquired for polyolefins coming
from building & construction waste (B&CW), divided into 9 different density fractions, ranging
from <0.88 g/cm3 up to 0.96 g/cm3 and in different color classes.

“Rapid discrimination of plastic packaging ma terials using MIR spectroscopy coupled

with independent components analysis (ICA),” 2014 A. Kassouf, J. Maalouly, D. N. Rutledge, H.
Chebib, and V. Ducruet Plastic packaging wastes increased considerably in recent decades, raising
a major and serious public concern on political, economic and environmental levels. Dealing with
 10

this kind of problems is generally done by landfilling and energy recovery. However, these two
methods are becoming more and more expensive, hazardous to the public health and the
environment. Therefore, recycling is gaining worldwide consideration as a solution to decrease
the growing volume of plastic packaging wastes and simultaneously reduce the consumption of
oil required to produce virgin resin.

“Deep: convolutional neural network applies to face recognition in small and medium
databases,” 2018 M. Wang, Z. Wang, and J. Li, this paper proposes a method combining local
binary patterns (LBP) and deep convolution neural network. This paper extracts LBP features of
face image as an input of CNN, and train the CNN network with the LBP features, then use the
trained network for face recognition, so that we can get rid of disadvantages of poor stability of
CNN gray scale and identify the trained CNN network more effectively.

“Application of foreground object patterns analysis for event detection in an innovative

video surveil lance system,” 2015 D. Frejlichowski, K. Go´sciewska, P. Forczmanski, and ´ R.
Hofman SmartMonitor is an innovative surveillance system based on video content analysis. It is
a modular solution that can work in several predefined scenarios mainly concerned with
home/surrounding protection against unauthorized intrusion, supervision over ill person and crime
detection. Each scenario is associated with several actions and conditions, which imply the
utilization of algorithms with various input parameters.

“PET waste classification method and plastic waste database - WaDaBa,” 2018 J. Bobulski
and J. Piatkowski. The main purpose of this work was creation of a plastic waste database of
images of objects constituting the typical contents of municipal waste. This group of waste, by
using methods of Computer Vision can be automatically selected on the sorting lines businesses
for waste disposal. Digital images of items that will be received for processing should reflect the
specific conditions of places where real objects have to be found. Thus, each thing is placed in this
database should be presented in the course of several collections of images, taking into account
different lighting conditions and different arrangement relative to the image recorder, and the
different degree of deformation of these objects as a result of previous processes
 11

“Waste classification system using image processing and convolutional neural networks,”
2019 J. Bobulski and M. Kubanek Image segmentation and classification is more and more being
of interest for computer vision and machine learning researchers. Many systems on the rise need
accurate and efficient segmentation and recognition mechanisms. This demand coincides with the
increase of computational capabilities of modern computer architectures and more effective
algorithms for image recognition.

“CNN use for plastic garbage classification method,” 2019 J. Bobulski and M. Kubanek
Image segmentation and classification is more and more being of interest for computer vision and
machine learning researchers. Many systems on the rise need accurate and efficient segmentation
and recognition mechanisms. This demand coincides with the increase of computational
capabilities of modern computer architectures and more effective algorithms for image
recognition. The use of convolutional neural networks for the image classification and recognition
allows building systems that enable automation in many industries.

2.1 EXISTING SYSTEM

Plastic dataset is considered to make the dataset segmentation, include the CNN model
perform in the dataset. The remaining steps after the segmentation the particular region are taken
in the dataset. That regions are covered by mask it detects plastic. And also, we classify the plastic
into PS, PP, PVC, PET, LDPE, HDPE and other. A report on the impact of plastic pollution, one
of the first to document the impact of discarded plastic on the health of the poorest people in the
world, estimates that between 400,000 and one million people die every year because of diseases
and accidents linked to poorly managed waste in developing countries.

Additional developments and comprehensive attempts should be carried out in the field of
skin disease recognition in the future. The following summarizes four possible directions for
research development. Deep learning has good generalization ability in Plastic recognition;
however, it requires many data in the learning process. Few data may cause insufficient feature
extraction, which will affect the diagnosis and recognition of lesions. One of the significant
 12

problems in Plastic image recognition is the difficulty in obtaining a large number of data sets due
to the complications in Plastic waste image collection.

2.2 DRAWBACKS OF EXISTING SYSTEM

• Difficult to Monitor Performance - It is not easy for managers to monitor their staff’s
progress and performance without them being in the same office space.
• To avoid all these limitations and make the working more accurately the system needs to
be computerized in a better way.
• Consumes large volume of paper work. Needs manual calculations.
• This is especially escalated if the job role requires a lot of “background duties” that can’t
be monitored on a work’s system. Financial burden on the world, Morbidity and mortality
Social and mental distance between people.

2.3 PROPOSED SYSTEM

When designing the CNN model for plastic type classification into 7 different types such
as HDPE, PET, LDPE, PVC, PP, PS, other. The first step is collection of appropriate images
of plastics and next step involves in renaming of these images. After that we need to fix the
size of the input image. At time the high resolution of images may lead to overloading of the
computational units and increase the memory. At same time the lower resolute image makes
the identification difficult.

Due to its poor quality the object identification process may be affected and decease the
performance of the model. We can use the image pixel width * height with 64*64 and 120 *
120 pixels. After this we need to decide the number of layers to be used inside the CNN model.
We analysed the model with different approaches. First, we used LeNet-5 in introduced by
Yann LeCun, LeNet-5 has 7 layers in which input layer is excluded. The LeNet-5 input layer
is of 32*32 image size. Secondly, we used the AlexNet (Figure 3) structure. In this the first
layer consisted of 64 filter of size 11*11.

In our proposed work, the simplified model for skin cancer identification is used. In this
model the preparation of the input image for validation and learning is the important phase. In
 13

addition to convolutional layer, we have use max-pooling layer, activation function, drop-out
layer, SoftMaxlayer. The convolutional layer extracts the feature and passes it to the next layer.
The system as a whole is examined, and the system's inputs are recognised. The different
procedures are linked to the organisations' outputs. The goal of system analysis is to become
aware of the problem, identify the important and decisional variables, analyse and synthesise
the numerous components, and come up with an optimal or at least adequate solution or plan
of action.

2.4 ADVANTAGES OF PROPOSED SYSTEM

• Avoiding local storage of data.

• By reducing the costs of storage, maintenance and personnel.
• It reduces the chance of losing data by hardware failures.
• Life style modifications
• Health awareness
• Importance of health
• Advantages of real-time and fast and are suitable for data analysis of a large number of
people.
• The sensitivity, spatial resolution and accuracy of its prediction result is improved.
 14

CHAPTER 3

SYSTEM SPECIFICATION

3.1 HARDWARE REQUIREMENTS

• Processor Type : Pentiumi3

• Speed : 3.40GHZ
• RAM : 4GB DD2RAM
• Hard disk : 500GB
• Keyboard : 101/102 Standard Keys
• Mouse : Optical Mouse

3.2 SOFTWARE REQUIREMENTS

• Operating System : Windows10

• Front end : Google-COLAB
• Coding Language : Python
 15

CHAPTER 4

SOFTWARE DESCRIPTION

4.1 FRONT END PYTHON

Python is a general-purpose interpreted, interactive, object-oriented, and high-level

programming language. Python source code is also available under the GNU General Public
License (GPL). It provides constructs that enable clear programming on both small and large
scales. It features a dynamic type system and automatic memory management, supports multiple
programming paradigms, including object- oriented, imperative, functional and procedural, and
has a large and comprehensive standard library. Python is open-source software and has a
community-based development model. Python and C Python are managed by the non-profit
Python Software Foundation.

4.2 FEATURES OF PYTHON

• Easy-to-learn − Python has few keywords, simple structure, and a clearly defined syntax.
This allows the developers to learn the language quickly.
• Easy-to-read − Python code is more clearly defined and visible to eyes.
• Easy-to-maintain − Python's source code is fairly easy to maintain.
• A broad standard library − Python's bulk of the library is very portable and cross-
platform compatible on UNIX, Windows, and Macintosh.

• Interactive Mode−Python has support for an interactive mode which allows interactive
testing and debugging of snippets of code.
 16

• Portable − Python can run on a wide variety of hardware platforms and has the same
interface on all platforms.

• Databases − Python provides interfaces major commercial databases.

• GUI Programming − Python supports GUI applications that can be created and ported
to many system calls, libraries and windows systems, such as Windows MFC, Macintosh,
and the X Window system of Unix.

• Scalable − Python provides a better structure and support for large.

4.3 GOOGLE COLAB

Google Co lab is a free and open source distribution of the Python and R programming
languages for data science and machine learning related applications. The Co lab is a free cloud
service based notebooks for machine learning, education and research.

Co lab Notebook is a web application that allows to create and share documents that contain

• Live code (e.g. Python code)

• Visualizations

• Explanatory text (Written in markdown syntax)

Co lab Notebook is great for the following use cases

• Learn and try out python

• Data processing / Transformation

• Long running background computation

• Statistical Modelling

• Machine Learning
 17

Applications of Google Co-lab Notebook

Co-lab is used extensively in the machine learning community with applications including
Getting started with TensorFlow, developing and training neural networks, experimenting with
TPUs, disseminating AI research, creating tutorials. A very attractive feature that Google offers to
the developers is the use of GPU. Google Co-lab is popular because of the support of GPU and
that is also totally free. The main aim behind the free Google co-laboratory for the public to make
the software a standard in the academics for teaching machine learning and data science purpose.
It may also have hidden long term perspective of building a customer base for Google Cloud APIs
which are sold per-use basis. Co lab allows to write and execute arbitrary python code through the
browser and is especially well suited to machine learning, data analysis and education.

Advantages

Sharing

Sharing of Google Colab notebooks is very easy. Everyone with a Google account can just
copy the notebook on his own Google Drive account. No need to install any modules to run any
code, modules come preinstalled within Google Colab.

Versioning

Saving the notebook to Github is done in just one simple click on a button. No need to
write “git add git commit git push git pull” codes in your command client

Code Snippets
Google Colab has a great collection of snippets you can just plug in on the code.
 18

Forms for Non-Technical Users

Not only programmers have to analyze data and Python can be useful for almost everyone
in an office job. Th e problem is non-technical people are scared to death of making even the tiniest
change to the code. But Google Colab has the solution for that. Just insert the comment #@param
{type:”string”} and you turn any variable field in an easy-to-use form input field.

Performance

Use the computing power of the Google servers instead of the own machine Running
python scripts requires often a lot of computing power and can take time. By running scripts in the
cloud, you don’t need to worry. The local machine performance won’t drop while executing your
Python scripts.
 19

CHAPTER 5

MODULE DESCRIPTION

5.1 PROBLEM DEFINITION

The image recognition method for plastic detection and classification based on machine
learning is to extract the features of plastic waste by manually setting extractors and classifying
them by traditional machine learning methods. This method requires great professional
knowledge to conduct deep exploratory data analysis and reduce its dimension. Finally, after a
complex parameter adjustment, the results can be outputted, which requires a large amount of
time and energy. This method has low portability, so the feature engineering is effective only in
the same field. These shortcomings limit the development of machine learning in plastic waste
recognition. Deep learning has achieved good results in image recognition with its advancement.
The deep learning method can automatically mine the deep-seated nonlinear relationship in
plastic waste images and do not need to establish feature engineering compared with the
traditional image recognition methods. The extraction efficiency is also efficient. Deep learning
is adaptable and easy to transform, and the technology can be more easily adapted to different
fields and applications.

5.2 OVERVIEW OF THE PROJECT

The process used to find out consumer waste can be recycled by combining the
secondary use of raw materials with a change in their condition and composition The
fundamental issue addressed in this article is the development of an automatic technique for
sorting plastic waste. This necessitates sorting waste into fractions other than metal, bio, plastic,
paper, and glass. Because not all of the components in each group are acceptable for reuse today,
advanced procedures must be used to distinguish the type of material in each group This can
 20

sort into the four categories listed above, PS, PP, PE-HD, and PET, and can be used in a sorting
factory or at home by residents. Plastic categorization by hand is a time-consuming and costly
operation.

5.3 MODULES DESCRIPTION

In this project, contains the following modules

• Dataset
• Applied Plastic Waste Detection Field
• Image Preprocessing
• Image Data Augmentation
• Evaluating Indicator

5.3.1 Dataset

Data transformation is the process of to fulfil the demands of a deep learning model,
information is transformed from one kind and the structure to another. Size adjustment,
normalisation, and greying are the most widely used data conversion technologies according to
the literature. The type of the plastic, the surroundings, the equipment, and the lighting conditions
all have an impact on the image quality. The recognition effect will be harmed by a low-quality
image, as a result, there is a loss of precision and a higher cost of calculation. Some of the most
commonly used approaches are spatial domain filtering, change detection filtration, and solving
differential equations.

5.3.2 Applied plastic waste detection Field

Prior to model training, image pre-processing is performed. The main goal is to remove
unnecessary information from the picture, increase the detectability of important and important
details hidden in the image, and substantially decrease the data to enhance the feature extraction
of the model. The fusion findings' ultimate output is achieved. Another option is to combine the
characteristics learnt from various models and then finalize findings of the Process flow in 2
 21

fusion procedures previously discussed. There are fourteen pieces of literature regarding multi
model fusion that have been gathered.

5.3.3 Image Preprocessing

In deep learning image recognition, a deep learning model has high requirements for
image quality because a good image quality can improve the model's generalization ability. Image
preprocessing is carried out before model training. The final output of the fusion results is
obtained. The other way is to fuse the features learned from multiple models to fuse the features
and then output the final results. The flow chart of the above-mentioned two fusion methods is
showed Fourteen pieces of literature are about multi model fusion collected.

5.3.4 Image Data Augmentation

Plastic Waste data are difficult to collect due to the problems of personal privacy and
professional equipment involved in the collection process of the plastic waste dataset.
Accordingly, less plastic waste data has been collected. Some plastic waste rarity makes the data
collection of this category less, resulting in the uneven distribution of the collected datasets. In
deep learning, small-scale datasets can easily lead to insufficient model learning and over fitting.

The solving of the shortcomings of the traditional plastic waste diagnostic process and
image recognition technology of plastic waste based on machine learning. Image recognition
based on machine learning is an interdisciplinary field integrating plastic waste imaging,
mathematical modelling, and computer technology through feature engineering and machine
learning classification algorithms to complete the recognition.

5.3.5 Evaluating Indicator

To evaluate the performance of deep learning in plastic waste image recognition, several
performance indicators are Few studies have been conducted on the pictures taken in real life.
The location of the plastic type is difficult to identify due to several problems, such as low pixels
and presence of shadows, and the plastic type area only occupies a small part or real scenes
pictures. This aspect needs further study. In clinical recognition models, attention mechanisms or
target detection methods can be introduced to focus on the type of plastic.
 22

The image recognition method for plastic detection based on machine learning is to extract
the features of plastic type by manually setting extractors and classifying them by traditional
machine learning methods. This method requires great professional knowledge to conduct deep
exploratory data analysis and reduce its dimension. Finally, after a complex parameter
adjustment, the results can be outputted, which requires a large amount of time and energy. This
method has low portability, so the feature engineering is effective only in the same field.

Figure 5.1 Dataset of Plastics

 23

Figure 5.5 Prediction Flow Diagram

 24

CHAPTER 6

RESULTS AND DISCUSSION

Solid wastes are being a main threat in environmental protection. In which plastics are
being a serious threat. Since its production rate is enormously high and reusing of these plastics
are low. Plastic categorization by hand is a time-consuming and costly operation. This necessitates
sorting waste into fractions other than metal, bio, plastic, paper, and glass. Thus, our Model
predicts the plastic and not a plastic image and classify the images into PS, PP, PE-HD, and PET.
This is an energy conserving technique. We got a performance metrics of accuracy 92 percent for
our classification model.

Figure 6.1 Output of Classification

 25

CHAPTER 7

CONCLUSION

Various factors in the environment result in the transformation of model concepts

and change of performance. The factors that drive the model should be understood to make
certain decisions. Currently, the interpretability research on plastic waste identification
models is a big gap in this field and must be explored. In the plastic type recognition field,
deep learning models can be applied to identify plastic. However, these algorithms are still
a ``black box'' to generate predictions on the basis of input data.

There is no specify interpretation on what plastic waste detection features does the
deep learning model as a basis for judgment. The development of deep learning should be
trustworthy and explainable. Making the algorithm public and transparent in decision-
making will give users reliability and security. Interpretability research on plastic waste
detection can resolve prejudices and auditing brought about by artificial intelligence
Interpretability makes artificial intelligence open and transparent in legal, moral, and
philosophical aspects.
 26

APPENDIX

SAMPLE CODING
# Part 1 - Building the CNN
import tensorflow as tf
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from keras.layers import Dropout
# Initialising the CNN

classifier =tf.keras.Sequential()
input_shape_x=127
input_shape_y=127
#(X_train, y_train), (X_test, y_test) = "/content/drive/MyDrive/classification/train"
classifier.add(tf.keras.layers.Conv2D(16, (3,3), input_shape = (input_shape_x,
input_shape_y, 3), activation = 'relu'))

classifier.add(tf.keras.layers.MaxPooling2D(pool_size = (2, 2)))

classifier.add(tf.keras.layers.Conv2D(32, (3, 3), activation = 'relu'))
classifier.add(tf.keras.layers.MaxPooling2D(pool_size = (2, 2)))
classifier.add(tf.keras.layers.Conv2D(64, (3, 3), activation = 'relu'))
classifier.add(tf.keras.layers.MaxPooling2D(pool_size = (2, 2)))

# Step 3 - Flattening
classifier.add(tf.keras.layers.Flatten())
# Step 4 - Full connection node

classifier.add(tf.keras.layers.Dense(units = 256, activation = 'relu'))

classifier.add(tf.keras.layers.Dropout(0.1))
classifier.add(tf.keras.layers.Dense(units = 64, activation = 'relu'))
#importany this retrives most info and ignores unecessary things

# classifier.add(tf.keras.layers.Dense(43, activation="softmax"))
classifier.add(tf.keras.layers.Dense(units = 7, activation = 'softmax'))
# Compiling the CNN
 27

classifier.compile(loss="categorical_crossentropy", optimizer="adam",
metrics=["accuracy"])

# Part 2 - Fitting the CNN to the images

#classifier.compile(optimizer = 'adam', loss = 'categorical_crossentropy', metrics =

['acc'])

# Part 2 - Fitting the CNN to the images

from keras.preprocessing.image import ImageDataGenerator

train_datagen = ImageDataGenerator(rescale = 1./255,

shear_range = 0.2,
zoom_range = 0.2,
horizontal_flip = True)

test_datagen = ImageDataGenerator(rescale = 1./255)

training_set =
train_datagen.flow_from_directory('/content/drive/MyDrive/classification/train',
target_size = (input_shape_x, input_shape_y),
batch_size = 4,
class_mode = 'categorical')#binary

test_set = test_datagen.flow_from_directory('/content/drive/MyDrive/classification/test',
target_size = (input_shape_x,input_shape_y),
batch_size = 8,
class_mode = 'categorical')
#binary
epochs = 15
# history = classifier.fit(training_set, epochs=epochs, batch_size=64,
validation_data=(test_set))

history=classifier.fit(training_set,
steps_per_epoch =153,
epochs = 180,
validation_data = test_set,
validation_steps = 58)
 28

REFERENCES

[1] Solid waste issue: Sources, composition, disposal, recycling, and valorization” 2018
By panelHussein I.Abdel-ShafyaMona S.M.Mansou.
[2] A review on au tomated sorting of source-separated municipal solid waste for
recycling” 2016 By Sathish Paulraj Gundupalli 1, Subrata Hait 2, Atul Thakur 3.
[3] Influence of shape and size of the particles on jigging separation of plastics mixture,”
2015 Fernando Pita 1, Ana Castilho 2.
[4] Intelligent solid waste processing using optical sensor based sorting technology,” 2010
Jiu Huang; Thomas Pretz; Zhengfu Bian All Authors.

[5] Advanced waste-splitting by sensor based sorting on the example of the MTPlant
oberlaa,” 2020 Janusz Bobulskia and Mariusz Kubanekb .
[6] Upgrading the quality of mixed recycled aggregates from construction and demolition
waste by using near-infrared sorting technology,” 2015 InigoVegas,Kris Broosb,Peter
Nielsenb, Oliver Lambertz,Amaia Lisbonaa.
[7] Case study: interpretability of fuzzy systems applied to nonlinear modelling and
control,” 2017 Krzysztof Cpalka.
[8] Real-time hyperspectral processing for automatic nonferrous material sorting,”
2012Artzai Picon, Aranzazu Bereciartua, Jone Echazarra, Ovidiu Ghita, Paul F.
Whelan, Pedro M. Iriondo.
[9] Industrial application for inline material sorting using hyperspectral imaging in the
NIR range,” 2005 Petra Tatzer,Markus Wolf,Thomas Panner.
[10] Sorting of polypropylene resins by color in MSW using visible reflectance
spectroscopy,” 2010 S. M. Safavi, H. Masoumi, S. S. Mirian, and M. Tabrizchi.
[11] utilization of hyperspectral imaging for impurities detection in secondary plastics,”
2012 S. Serranti, A. Gargiulo, G. Bonifazi, A. Toldy, S. Patachia, and R. Buican.
[12] Characterization of post-consumer polyolefin wastes by hyperspectral imaging for
quality control in recycling processes,” 2011 S. Serranti, A. Gargiulo, and G. Bonifazi.
[13] Classification of polyolefins from building and construction waste using NIR
hyperspectral imaging system,” 2012 S. Serranti, A. Gargiulo, and G. Bonifazi.
 29

[14] Rapid discrimination of plastic packaging ma terials using MIR spectroscopy coupled
with independent components analysis (ICA),” 2014 A. Kassouf, J. Maalouly, D. N.
Rutledge, H. Chebib, and V. Ducruet.
[15] Deep: convolutional neural network applies to face recognition in small and medium
databases,” 2018 M. Wang, Z. Wang, and J. Li.
[16] Application of foreground object patterns analysis for event detection in an innovative
video surveil lance system,” 2015 D. Frejlichowski, K. Go´sciewska, P. Forczmanski,
and ´ R. Hofman.
[17] PET waste classification method and plastic waste database - WaDaBa,” 2018 J.
Bobulski and J. Piatkowski.
[18] Waste classification system using image processing and convolutional neural
networks,” 2019 J. Bobulski and M. Kubanek.
[19] CNN use for plastic garbage classification method,” 2019 J. Bobulski and M.
Kubanek.

View publication stats