Week 2— Waste Classification

INTRODUCTION

Hi everyone,we are Hasan AKALP ,Umut PİRİ & Dilara İŞERİ. For our semester project in Machine Learning Course we are trying to waste classification.

This week we propose you the related works in the literature. Because these related works will give you an idea about what we want to do.Continue to read our article to better understand the project we are trying to do and to have more information about the subject.

As you can see we mentioned about the possible methods in the literature. Now let’s explain them detaily!

We mentioned the data sets we will use in our previous blog. We should remind you again. We will basically use two different data sets for this project. The first data set we will use is Gary Thung’s TrashNet dataset. We would like to touch on the features of this data set.

The dataset spans six classes: glass, paper, cardboard, plastic, metal, and trash. Currently, the dataset consists of 2527 images:

• 501 glass
• 594 paper
• 403 cardboard
• 482 plastic
• 410 metal
• 137 trash

The pictures were taken by placing the object on a white posterboard and using sunlight and/or room lighting. The pictures have been resized down to 512 x 384, which can be changed in data/constants.py

Another dataset is Sashaank Sekar’s Waste Classification dataset. This dataset spans two classes: organic, and inorganic. The dataset has 22564 train samples and 2513 test samples.

After briefly recalling our data sets, let’s examine related works

ARTICLE -1-

Fine-Tuning Models Comparisons on Garbage Classification for Recyclability

In this work, it is aimed to develop a deep learning application which detects types of waste into trash in order to provide recyclability with vision system. Training and testing will be performed with image data consisting of several classes on different waste types. The data set used during training and testing will be generated from original frames taken from waste images. The data set used for deep learning structures has a total of 2527 images with 6 different classes.

MATHERIALS & METHODS USED

1-Deep Learning

2-Convolutional Neural Network

3-Softmax

4-Support Vector Machine

5-Dataset

6-Proposed Method

ARTICLE-2

Classification of Trash for Recyclability Status

Again, the aim of this study is to classify waste images according to various classes using certain methods.

In this study, two methods are generally used:

• Convolutional Neural Network
• Support Vector Machine

The results of the classification process with these two methods are as follows.

The SVM achieved better results than the CNN. It achieved a test accuracy of 63%. The training error was 30%. The SVM is a relatively simpler algorithm than the CNN, which may attribute to its success in this task.

Convolutional Neural Network As stated in the experiments section, they had trouble with training the network. The network seemed to not learn, as the test accuracy we achieved in the experiment described was only 22%. This is barely.

ARTICLE-3

RecycleNet: Intelligent Waste Sorting Using Deep Neural Networks

The aim of this study is to classify wastes for waste management and recycling.Or as they say, “The main purpose of this work is to demonstrate an effective intelligence system to classify selected classes of common garbage materials.”

Their model RecycleNet is carefully optimized deepconvolutional neural network architecture for classification ofselected recyclable object classes. This model decreased the number of parameters in a 121 layered network from 7 millionto about 3 million

MODELS AND METHODS

• Deep Residual Networks
• MobileNet
• Inception-ResNet, Inception-v4 (InceptionResNetV2)
• Densely Connected Convolutional Networks
• RecycleNet

ARTICLE-4

Smart Trash Net: Waste Localization and Classification

The aim of this project is to increase the recycling rate by automating garbage classification. Given an image of jumbled trash / waste which contains two or more different pieces of waste, they want to localize the image and classify the different forms of waste into three categories: recyclable, paper, and landfill.

MODELS AND METHODS

• Faster R-CNN

In this week’s blog, we briefly talked about the projects and the methods they used in the literature about the project. Thank you for reading this blog. See you next week.