MSc Cloud Computing: Presentation on Waste Segregation using DNN

This presentation explores the use of Deep Neural Networks (DNN), specifically Convolutional Neural Networks (CNN), for automated waste segregation, focusing on food waste. The research question investigates the feasibility of creating an automated architecture for waste segregation from food joints. The objectives include identifying and categorizing food waste based on images to generate nutritional data. The presentation details the DNN architecture, which utilizes pixel-based data input and comparison with datasets through pooling and gradient descent algorithms. The methodology involves storing large datasets for training the system to identify food waste, with CNN enabling food object detection and recognition. The presentation concludes that the research objectives align with the problem, highlighting how food detection and identification can facilitate waste segregation. The deep learning solution uses a Convolutional Neural Network, with techniques detailed in the presentation. This presentation, contributed by a student, is available on Desklib, a platform providing AI-based study tools.

A Presentation on Waste
Segregation using DNN

Research Question
 How creation of automated architecture for bargaining on waste
segregation is doable
 If Segregation of wastes from food joints can be automated using Deep
Neural Network
 What aspect of Deep Neural Network can help in this automation
 Role of Convoluted Neural Network in success of the architecture

Research Objectives
 The machine learning solution is to address a vital issue of the society
which is wastage of left over food
 Administrating wastage of food is a priority as this can lead to several
environmental as well as economic benefits for the society
 To address the imbalance in availability of food resulting in malnutrition
of hundreds in poorer countries
 The solution needs to identify and categorize the contents of food and
generate the nutritional data of wasted food (Orhorhoro, Ebunilo and
Sadjere 2017)
 To be able to distinguish wasted food based on their images

Design and Methodology
The Deep Neural Networking solution
is based on CNN architecture
Convolutional Neural Network or CNN
is used to enable the solution to
segregate food visually
Pixel based data is taken as input and
compared with data sets through
pooling
Gradient Descent based algorithms are
used to get towards accurate
classification of image
The activations are then collected and
flattened through matrix
multiplications in the fully connected
layer
Then the output gets generated based
on distinctions made from the process

Methods and Technologies
of the DNN Architecture
To process the images and accurately
identify food large sets of data is to be
stored in the database.
These data are for training the system
to identify and segregate food waste
CNN lets the system detect and
identify food objects
The food image is compared with that
of objects present in the database in
pooling layers (Yanai and Kawano
2015)
Food detection and recognitions are
done by SVM as classifier and BoF
model
Higher recognitions in confusion
matrix denote availability of higher
training data

Conclusions
 From the above slides it can be concluded that the research objectives
are in line with the problem in research question
 The objectives describe how detection and identification of food can
help segregate food waste
 The deep learning solution for detection and identification is explained
in the next slides
 The deep learning architecture used is Convolutional Neural Network
and techniques used by this solution are mentioned

References
 Orhorhoro, E.K., Ebunilo, P.O. and Sadjere, E.G., 2017. Development of a
Predictive Model for Biogas Yield Using Artificial Neural Networks (ANNs)
Approach. American Journal of Energy and Power Engineering, 4(6),
pp.71-77.
 Yanai, K. and Kawano, Y., 2015, June. Food image recognition using deep
convolutional network with pre-training and fine-tuning. In 2015 IEEE
International Conference on Multimedia & Expo Workshops (ICMEW) (pp.
1-6). IEEE.