Scribd
Upload
42 views18 pages

Thesis Research Deep Learning

The document is an undergraduate thesis by Arla Zeqaj focusing on deep learning and image processing, detailing concepts such as neural networks, image segmentation, and object detection versus recognition. It discusses the applications of deep learning in various fields, including healthcare and autonomous vehicles, as well as challenges associated with image datasets. The thesis also highlights the structure and function of convolutional neural networks in image recognition tasks.

Uploaded by

azeqaj22
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
42 views18 pages

Thesis Research Deep Learning

The document is an undergraduate thesis by Arla Zeqaj focusing on deep learning and image processing, detailing concepts such as neural networks, image segmentation, and object detection versus recognition. It discusses the applications of deep learning in various fields, including healthcare and autonomous vehicles, as well as challenges associated with image datasets. The thesis also highlights the structure and function of convolutional neural networks in image recognition tasks.

Uploaded by

azeqaj22
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 18

Deep learning &

Image processing

Undergraduate Thesis
research Arla Zeqaj
01 - Deep learning
02 - Image Processing
03 - Image Segmentation
04 - Object Detection
vs Recognition
05 - Image Datasets
06 - Neural Networks
07 - Convolutional
Neural Networks
01 - What is deep learning?

Deep learning is a subset of machine learning


and artificial intelligence: a technique of
training computers to mimic the way neurons in
the human brain process and learn information.
Each layer processes information from the
previous layer and passes it on to the next,
gradually extracting higher-level features and
patterns.
(Deep Learning: The Mechanics of Magic, n.d.,
para. 1,5)
Deep learning is redefining the very way we
experience life.
Examples:
Computer vision: unlocking your phone with just a
glance
Natural language processing: Google Translate, voice
assistants, models like GPT (Generative Pre-trained
Transformer) craft human-like conversations
Healthcare: identifying tumours in medical scans
Autonomous vehicles: Self-driving cars
(Deep Learning: The Mechanics of Magic, n.d., sec. 4)
02 - Image Processing techniques

Image processing is a method which is used to


improve raw images and to abstract valuable
information from them.

Digital image processing uses many techniques like:

image segmentation
image enhancement
image restoration
image acquisition
classification & description
image representation
image compression

Rani, Neetu. (2017).


03 - Image segmentation

The process of dividing an image into


different regions based on the
characteristics of pixels to identify
objects or boundaries to simplify an
image and more efficiently analyze it

(Image Segmentation, n.d.)


Segmentation of an image example

(Image Segmentation, n.d.)


Segmentation impacts a number of domains:

the filmmaking industry: the software behind green


screens to crop out the foreground

in satellite images: to track objects in a sequence


of images and to classify terrains, like petroleum
reserves

medical applications: the identification of injured


muscle, the measurement of bone and tissue, and the
detection of suspicious structures to aid
radiologists

(Image Segmentation, n.d.)


04 - Object Detection vs Recognition

can handle multiple


finding and locating
objects of different
objects of interest
Detection in an image or video
types and sizes in a
single image

outputs a label or a
identifying and score that indicates the
Recognition classifying objects class or the confidence
in an image or video of the recognition

(What Is the Difference Between Object Detection and


Object Recognition?, 2024)
Obj detection Obj recognition

faces detection that face recognition of a


are present in a specific person
photo

pedestrian detection handwritten digit


recognition
traffic sign
detection semantic segmentation

(What Is the Difference Between Object Detection and


Object Recognition?, 2024)
05 - Features & challenges in Image Datasets

Example: Cdiscount.com image dataset


More than 15 million images at 180x180 resolution
More than 5000 categories

Challenges
1.Huge Dataset (~100GB+), requiring significant disk
space and memory
2.Same products containing up to 4 images
3.High Intra-Class Variability: multiple
subcategories
4.Noisy and Low-Quality Images

(CDiscount’s Image Classification Challenge, n.d.)


05 - Features in Image Datasets

Pixel Data Array Resolution & Dimensions

Color & Texture Features Shapes & Contours


05 - Challenges in Image Datasets

Data Diversity: variability in lighting, angles, background


Data Annotation: annotating with labels is labor-intensive
Data Bias: biased sources, annotators, collection methods
Data Privacy: images of people or sensitive information

(Solution, 2023)
06 - What is a Neural Network?

Neural networks are computing systems with


interconnected nodes that work much like neurons in
the human brain.

They can recognize hidden patterns and correlations


in raw data, cluster and classify it, and – over time
– continuously learn and improve.

(Neural Networks: What Are They and Why Do They


Matter?, n.d.)
06 - ANNs in images
Neural networks for image recognition typically consist of
several layers of neurons that process the image in a
hierarchical manner.

receive raw
pixel
values

(Vungarala, 2023)
07 - Convolutional Neural Network components

CNNs contain five types of layers:


input, convolution, pooling, fully connected and
output.
Each layer has a specific purpose, like summarizing,
connecting or activating. They have popularized image
classification and object detection.

(Neural Networks: What Are They and Why Do They


Matter?, n.d.)
References
Deep learning: The mechanics of magic. (n.d.). ISO.
https://www.iso.org/artificial-intelligence/deep-learning

Image segmentation. (n.d.). Stanford Artificial Intelligence Laboratory.


https://ai.stanford.edu/~syyeung/cvweb/tutorial3.html

Rani, Neetu. (2017). Image Processing Techniques: A Review. Journal on


Today's Ideas - Tomorrow's Technologies. 5. 40-49.
10.15415/jotitt.2017.51003.

What is the difference between object detection and object recognition?


(2024, March 7). https://www.linkedin.com/advice/1/what-difference-
between-object-detection-recognition-ozwrf
References
CDiscount’s image classification challenge. (n.d.). Kaggle.
https://www.kaggle.com/c/cdiscount-image-classification-
challenge/overview

Solution, G. T. (2023, November 6). Challenges and solutions in Image


Data Collection for Machine Learning. Medium.
https://medium.com/@aanchalgts.ai/challenges-and-solutions-in-image-data-
collection-for-machine-learning-195b66f6d1b5

Neural Networks: What are they and why do they matter? (n.d.). SAS.
https://www.sas.com/en_us/insights/analytics/neural-networks.html

Vungarala, S. K. (2023, May 1). Image Recognition — How neural network


identifies? - Seshu Kumar Vungarala - Medium. Medium.
https://medium.com/@seshu8hachi/image-recognition-how-neural-network-
identifies-17ec0ccda662

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy