International Journal of Electronics and Microcircuits 2022; 2(1): 51-56

E-ISSN: 2708-4507
P-ISSN: 2708-4493
IJEM 2022; 2(1): 51-56 Image recognition with machine learning
© 2022 IJEM
www.microcircuitsjournal.com
Received: 13-01-2022 Parth Kumar Thakur
Accepted: 16-02-2022

Parth Kumar Thakur Abstract

School of Computer Science Image recognition or computer vision is a method that allows machines and computers to parse and
and Engineering, Vellore piece together relevant information from images. It is a fast-growing field of AI with many real-world
Institute of Technology, applications in robotics, artificial intelligence, self-driving car, surgical tool and much more. Computer
Chennai, Tamil Nadu, India vision can also be used to implement and render 3D models from 2D images and automating the
rendering process. Massive leaps have been made in the technology allowing for AI to implement real
world images within the software. A subset of this is seen in self driving cars, and leading companies
like Tesla which allow for quick and efficient image processing in both visual and non-visual mediums
like infrared to allow for the cars to navigate through their surroundings on their own. Even with the
growth Artificial Intelligences (AI) are far from perfect and are as of yet incapable of beating human
cognition, but that day may not be too far away before it arrives. Technological improvements have
already made self-driving cars a reality, propagating the value of image recognition and computer.
Probes to interplanetary missions will be requiring self-governing AIs due to the lacking capability of
instantaneous communication, which makes the field all the more relevant for the future growth and
progression of humanity.

Keywords: Facial recognition, machine learning, edge detection, pattern recognition, deep fake

1. Introduction
Machine learning has progressed significantly over the past few decades, overcoming many
challenges and becoming capable enough to recreate complicated 3D structures like the
human face accurately. The technology has only improved in recent years extending from
simplistic detection models working on objects to more complicated methods of analysing
visual input to recreate, enhance, modify and process the information being presented to the
systems. The capabilities of these algorithms and neural networks have increased even
further when they’ve been provided the datasets formed from previous neural networks to
work with and refine their algorithms for better results. A subset of Deepfake images, videos
and other multimedia methods have arisen with the rise of this technology. Simultaneously, a
rise in Artificial Intelligence geared towards differentiating and detecting these deep fakes
have also seen a rise in the past few years. One big application of this technology can be seen
in obtaining underwater images and using neural networks to remove the water, allowing for
a clearer view [1]. Another application can be seen with the usage of multiple images to
create a 3D render and model of the given images, allowing for processing and data to be
collected with ease [3]. With the increasing demand for self-driving cars and the need for
safety, better resource management, traffic and decreased pollution, we need better AI
models capable of learning faster than ever [4].
Definition of Machine Learning as a term varies depending on the context but the basis of
the idea all include using images and processing them to obtain data. Machine learning with
security cameras, have allowed investigations to more easily detect and tell apart criminals,
with many security approaches using machine learning and facial recognition for
investigation and criminal identification [7]. Biometrics analysis and forensic rports have also
begun using the help of image recognition for selecting out clues and information that may
have otherwise been lost, or to process infrared imaging and ultrasonic mapping to obtain a
better and more thorough investigation of the crime scene. Last, perhaps the largest use of
machine learning has been in product inspection for objects such as microprocessors,
Correspondence automobiles and agricultural and food produce [8]. The use of machine learning has
Parth Kumar Thakur
School of Computer Science automated the inspection of such processes, especially with regards to small and easy to miss
and Engineering, Vellore errors present in microprocessors etc.
Institute of Technology, A large part of image processing relates to machine learning and the development of
Chennai, Tamil Nadu, India Artificial intelligence (AI).

~ 51 ~
International Journal of Electronics and Microcircuits www.microcircuitsjournal.com

Machine vision has become a more and more important set in comparison to the limitations present from real images
assets and component in the development of neural as shown in the following figure [6].
networks as it has turned out to be one of the easiest modes
of obtaining information that can be taught to a neural Table 1: Real vs Synthetic data set for facial recognition
network. The improved capability and application in algorithms
robotics alongside the various algorithms that can recreate Common NME Challenging NME Private FR 10%
faces, image, geographical locations, terrains, ecological CVPR' 17 3.67
and global mapping, art, and help enhance visual images, CVPR' 18 2.98 5.19 0.83
and video editing in towards modern world has made ICCV' 19 2.72 4.52 0.33
Machine Vision synonymous with Artificial intelligence. CVPR' 19 3.56 6.67 -
Neural networks are becoming increasingly more capable at ICCV' 19 3.19 6.87 -
processing and analysing images, with real time application CVPR'20 3.36 5.74 0.17
and analysis providing a new level of reach to the field. (real) 3.37 5.77 1.17
Many machines, robots, systems are already making use of (synthetic) 3.09 4.86 0.50
this new technology inducing rapid growth. From the largest
to the smallest sector, machine vision and artificial The data shows the results of synthetic training for facial
intelligence are helping enhance and improve development recognition as a faster and more efficient way to train new
and lifestyle all over the globe. The increasing popularity of algorithms. It allows for improved facial landmark
self-driving cars also serve to make traffic accidents rarer processing and removes any hindrances such as lighting and
and help reduce carbon emissions from vehicles as electric angle concerns due to the synthetic nature of the database.
cars become more prominent alongside them. Additionally, this approach also provides an increased
With vast applications and many uses that only keep control over the fidelity of the data and allows for a much
growing, there is in an increasing need for research and more variable dataset that can be exponentially generated,
understanding to create better neural networks and machine tweaked, modified and reiterated.
learning for image processing, and machine vision. To A similar approach can be seen in image recognition
better develop, understand and improve upon this performed by virtual simulations for self-driving cars [4].
technology. The same can be used as a training method for humans
which provides more safety for learning and beginner
2. Neural Network Training Models for Image drivers in handling the new and as of yet unfamiliar vehicle.
Processing Random generation of terrain and image recognition allows
Computer vision systems uses technology based on image for the creation of realistic models of real-world terrain to
recognition, primarily based upon cameras and other visual provide for these simulations, combined with physical
devices to obtain data. The machine learning algorithm then simulations.
processes the data with a feedback loop checking for the
accuracy of the data and then assigning feedback on the 3. Applications of Machine Vision and Image Processing
basis of the accuracy of the detection. The feedback is taken There are an increasing number of applications for computer
in by the algorithm and applied on a larger data set linked vision ranging from collecting data, processing information,
together through and processed through multiple algorithms facial recognition, landmark recognition, real time self-
and run through many iterations to enhance the accuracy of driving cars, modifying image processing and much more
[25-42]
the algorithm. There is an inevitable human part in the usage . In this section, some of the major applications of this
of this method where the human operators actively train the technology will be highlighted.
machine and provide it with feedback on the accuracy of its
recognition. The most popular means of providing the Facial recognition
dataset so far has been the modern Completely Automated Neural networks and AI play a massive role in Facial
Public Turing test to tell Computers and Humans Apart recognition, which forms one of the primary uses of image
(CAPTCHA) training methods [5]. Originally intended as a recognition. It’s a technology that has been rapidly
means of telling humans and bots apart, the image expanding in its usage, from biometrics and identification to
recognition method has become a way to provide an information and data analysis model and graphics rendering.
accurate data set and method to train modern Artificial Deep fake and complicated video editing has allowed for
Intelligences (AI) by providing human feedback. The extremely realistic renders and videos to be developed by
efficacy of this method will soon turn the current training an AI and feeding it enough source images about a
CAPTCHA methods of image recognition obsolete as person. Alongside the many practical usage this has also
machines would be just as capable as humans in telling developed concerned for abuse and misuse and has raised
apart objects, if not better. concerns on an individual’s privacy and security on the
Another method arising for training Neural networks to internet even more so than ever before. Facial recognition is
recognise and tell apart images, including better edge also, used in government and military Airports [9] as a means
recognition, segmentation and pixel counting has been the of keeping track of passengers and personnel that have
usage of other machine learning algorithms to train newer arrived and are departing. It is used for face id unlocks in
ones. Training machine learning for processing images modern smart phones [10], a widely used and popular method
based on synthetic dataset has opened a much vaster and of security but with concerns over how the same can
much more flexible dataset to train neural networks and sometimes be tricked using an image of a person which has
databases from. An application of such can be seen with led to be the development of infrared and 3D face mapping
machine learning algorithms used to recognises faces, and which develops a 3D face database of a person and cannot
reconstruct them performing better with the synthetic data be tricked using a flat image. The same technology can be

~ 52 ~
International Journal of Electronics and Microcircuits www.microcircuitsjournal.com

seen certain college classrooms to take attendance. It is used a wide application of image processing software combined
on social media platforms like Facebook [11]. Cameras and with VGR’s in examples such as the da Vinci surgical
other devices come with facial recognition software to take systems [17]. The development of machine vision in these
better images and for improved lens focus [6]. Facial sectors has made increasingly difficult procedures and
recognition databases also play a major role in crime terminal injuries less of a risk. Cancer treatment, and
investigation and modern-day law enforcement [7]. Facial delicate surgical procedures for neurological surgeries and
recognition is used in automobiles to prevent car theft and organ transplants have both greatly benefitted from the
increase security. There is increasing usage of Facial development of this technology as image processing and
recognition in marketing for better targeting ads and more machine vision allows the use of far more complex robotic
effective usage of resources [12]. surgical tools that can provide a better and more controlled
environment for the surgeons to work in.
Bar code reading
A major usage of image recognition is barcode reading and Satellite imaging and navigation
QR code reading where the images are used to encode a Imagine processing is used in forming maps across the
certain piece of data and can then be decoded by the globe using satellites. GPS and navigation systems can also
application reading the image. It is often used to store data rely on image processing to detect and locate various
on physical goods and products in grocery stores and other terrains. Real world views and 3D maps formed by google
physical appliances [13]. They form a way to store data showing each location on their maps are mapped using 360
physically on devices in a compact manner allowing people cameras set on vehicles which are then processed via image
to track specific devices, their dates of manufacture, stature, processing software to provide. Locating rescue spots,
location of origin and product descriptions with the usage of research locations, monitoring changes to the polar caps and
a barcode reader through any local system. This allows for a many other uses tie into this segment of image recognition.
far easier distribution and tracking method than its Things such as wildfires, flood and other natural disasters
predecessor and has wildly been spread out as a means of can be detected before they occur with the help of satellite
product management and tracking. The further evolution of imaging and image recognition. In [21], Weather prediction
the technology has come in the form of QR codes which models, detection and studies of climate change, increasing
uses a different format of image coding to store information. carbon emissions, sea levels rising also make use of image
recognition to predict these weather patterns. In [22],
Vision guided robots Hurricane and cyclone detections make use of Image
Vision guided robots (VGR) play an important role in recognition as well to improve their detection
multiple sectors with applications ranging from medical and methodologies and get predictions for harvesting rainfalls to
surgical robots that improve procedure and safety of the help farmers and the agriculture sector prepare accordingly
operation to drones and environmental scanning robots for for planting and harvesting their crops. This plays a vital
rescue operations. There is also a big role of visual guided role in the development and tracking of agricultural cycles
robots in military applications and guided missiles alongside and the economy of nations that depend heavily on the
anti-aircraft and missile security systems meant to take out weather for their agriculture. The same can be used to track
any unidentified aircrafts navigating the borders or entering the changing weather conditions across any given region for
a country’s aerospace without authorisation. They are also further climate change and weather studies.
used for inspection in the manufacturing and processing
industry [14]. This field also pertains to self-driving car, 4. Technologies of Machine Vision
otherwise known as automatic vehicles (AV) which has There are various technologies pertaining to machine vision
seen a massive increase in its popularity with large tech and image recognition [30-40]. With various methodologies
companies like Tesla spreading the technology and making and technologies meant for different subsets of image
it more and more feasible than ever before [4]. The same processing and recognition. The following list contains
technology is used in security and traffic cameras to detect some of the prominent methods and technologies used in
collision and inform medical or police services if required machine vision.
[18]
. Speeding tickets can also be charged and automatically
added based on the license plate of the speeding vehicle [19]. Image Stitching and Registering
Alongside the practical application, robotics and Image stitching is one of the most basic forms of machine
animatronics have developed massively with autonomous vision where multiple images are stitched together and
robots capable of mimicking conversations using vision overlapped to form a singular image. It can be used for
guidance to study their environments, the people higher resolution, for software implementations of focus and
participating in the conversation and appropriate responses zoom in cameras and for a wider-angle view using multiple
to any given situation. Other robots can help speed up images. The technology uses algorithm to determine image
manufacturing processes and reduce faulty products from alignment by relating coordinates of pixel from one to
being produced by relying on machine vision and guidance another to form a rough guideline of positioning and angle.
to analyse and maintain quality for the parts being produced. Image Registering then can use a set of key point and
matching features to minimize the errors and differences
Medical Usage that may crop up from the combination of multiple images.
Image processing can be used for X-rays, ultrasounds, The most common method used for this purpose is Random
chemotherapy, cancer detection and guided application and Sample Consensus (RANSAC) which is an iterative method
cancer cell extermination [15, 16]. It can be used to treat that uses mathematical models based on a given dataset of
cancer, tumour and help with a far more targeted approach observed points. This technology then further has models
towards treatment of many illnesses and conditions. There is for calibration whose purpose is to minimize the differences

~ 53 ~
International Journal of Electronics and Microcircuits www.microcircuitsjournal.com

between a camera lens image and the image formed using it down into smaller and smaller components interconnected
the image stitching and registering method used in this and tied together [24]. Some of the primary methods of blob
machine vision technology. The algorithm would observe detection are Laplacian of Gaussian (LoG), Difference of
distortion and exposure differences before correcting and Gaussian (DoG), Determinant of Hessian (DoH) and
optimising the image to be uniform with the overall Connected Components Labelling.
composition. Another method for image stitching is image
blending, which takes image comparison information from 5. Conclusion and Future Remarks
the calibration stage and uses the given information to In this paper we analysed the usage of Artificial
combine and adjust the image for an output projection. Intelligence, Neural Networks and Machine Learning for the
development of machine vision, and examined the various
Edge Detection methodologies, models, approaches used in the technologies
Edge detection is a technology that combines mathematical that make use of this field of Image processing. We have
methods to detect edges, corves and lines. The technique shown the widely spread-out usage of image processing
simplifies the process of image detection by reducing the across multiple fields such as education, medicinal
amount of information that needs to be processed by development, surgical processes, quality maintenance,
converting a 3D image into its subsequent 2D representation security, national defence, military operations, robotics,
while still containing the needed structural composition to manufacturing and processing, agricultural development,
maintain the image’s integrity and information [23]. There geo satellite mapping, navigation, graphic rendering, 3D
are various different methods applied with edge detection modelling and many other fields, and we’ve shown the
with two primary categories, zero-crossing based and application of Machine learning, Artificial intelligence and
search-based. Search based applied a first order derivative to Neural networks in the development of this technology.
detect edges by a measure of edge strength. The zero- While widely prevalent, the technologies for machine vision
crossing method on the other hand applies a second order and neural networks still remains in its infancy. With better
derivative to find the edges in the image. Some of the masks technological advances being made each day and
for edge detection are Prewitt Operator, Sobel Operator, development being made in the field alongside increasing
Robinson Operator, Laplacian operator. The first of them is research, and capabilities of hardware devices to run and
used to detect horizontal and vertical edges. Sobel operator refine these algorithms, we are on the cusp of a world with
works quite similarly to Prewitt operator, detection both enhanced and quick image processing capabilities, and
horizontal and vertical edges. Robinson operator makes use robotics capable of navigating and understanding the world
of a mask, rotating it all over in 8 separate directions to get around them in a way that had not been possible up until
direction edges and vectors for all of them. Kirsch compass now. The development of this technology, which has
mask is another mask operator that can find edges in all become synonymous with artificial intelligence has come up
directions. Finally, the Laplacian operator is a second order with concerns and pushback from concerned individuals
derivative mask, which has two sections, positive Laplacian about the abuse of deep fakes and the ease with which these
and negative Laplacian. One big usage of edge detection is can be produced, but various articles and research papers
in sharpening images, where the more edges are found, the have highlighted the benefits far outweighing the
sharper the image becomes after the edges have been disadvantages. Image processing and Artificial Intelligence
applied. It can be used in adding many outlines and effects are technologies that carry significant impact on the
in photo editing that are based around the shapes and edges development of the future of humanity and mark a vital step
present in an image. in the growth of the field of artificial intelligence and
computer science. The developments made in just the last
Blob Detection decade have been astronomical and with more importance
Blob detection, otherwise known as Connected-component being given to the field, the research and growth has only
labelling (CCL), connected-component analysis (CCA), increased with time.
blob discovery or region extraction is an algorithm that In this paper, the models described above, proposed by
applies graph theory to analyse and process images. In it, various authors and research papers for the development of
subsets of uniquely identified groups of pixels and sections image processing and the approach with which to tackle
are labelled, giving it the name connected-component machine learning in regards to image processing has
labelling. It is used to scan, analyse and detect connected highlighted multiple directions for the development of this
regions in an image. The process works with binary images technology. The paper discusses the various approaches in
but colour images can also be analysed using this method. which machine learning algorithms can be trained and
Since images contain various forms of information, pre- refined using the varying nature of data sets and feedback
processing to filter the irrelevant information out before loops dependant on both human input and machine input.
using blob detection is required often times. Blob detection The paper has also highlighted the differences between the
locates and finds any large clusters of pixels, bright pixels, two approaches alongside their advantages and
set against a dark background. That is, a set of pixels that disadvantages. Various application and usage of the
can form a distinctive collection and can distinguish itself technology have been highlighted in the paper as well,
from the background of the image. Blob Detection is often alongside the processes and technologies that form the basis
used in pair with other image recognition systems, of image processing and the methods in which these
mechanisms and component and can work based on a algorithms analyse and process any given image. Taking in
variety of information and data points depending upon the account the various methods of development, the
nature of the image processing and detection system. There advantages and disadvantages of a synthetic dataset, the
are multiple usages of real time blob detection which can methods used for developing and training machine vision,
help in processing and understanding an image by breaking it’s varied and diverse usage methods alongside various

~ 54 ~
International Journal of Electronics and Microcircuits www.microcircuitsjournal.com

technologies present in it such as Image Stitching and https://www.researchgate.net/publication/236189439_

Registering, Edge Detection and Blob Detection, the author Machine_vision_system_A_tool_for_quality_inspectio
concludes as such: n_of_food_and_agricultural_products
There is an ever-increasing need for the development of 9. NimraKhan, Marina Efthymiou. DCU Business School,
image processing algorithms in today’s growing multimedia Dublin City University, Glasnevin, Dublin 9, Ireland.
world of technology and internet, with more and more The use of biometric technology at airports: The case of
avenues and technologies such as Computer-Generated customs and border protection (CBP)
Imagery (CGI) in movies and other industries relying more 10. Prof. Hassan, Soliman, Ph.D, Department of
and more on machine learning algorithms to refine and communication, Engineering, Mansoura University,
process vast quantities of image and video data. There is Egypt. Ahmed Saleh, Ph.D, Department of Computer,
also a need for increasing security and infrastructure to Engineering, Mansoura University, Egypt. Eman Fathi,
prevent fraud and scams using deepfakes and machine Department of communication, Engineering, Mansoura
generated images, faces, identities and other similar University, Egypt, 2013 Jul, 73(2).
applications in this new growing world of image processing 11. Yaniv Taigman Ming, Yang Marc’Aurelio Ranzato,
capabilities. A deeper look into the developmental processes Facebook AI Research Menlo Park, CA, USA. Lior
and a better understanding of today’s needs for the Wolf, Tel Aviv University, Tel Aviv, Israel. DeepFace:
technology is required. Closing the Gap to Human-Level Performance in Face
Verification.
6. Authorship Statement https://research.facebook.com/file/266870805034649/d
All persons who meet the authorship criteria have been eepface-closing-the-gap-to-human-level-performance-
listed as authors and took public responsibility for the in-face-verification.pdf
content including participation in the concept, analysis, 12. Senior Systems Engineer, Information Technology.
writing, and revision of the manuscript. Master of Science in IT, Wilmington University, New
Castle, Delaware, USA. Facial Recognition
7. Conflict of Interest Technology.https://www.researchgate.net/publication/3
The authors declare that they have no conflicts of interest. 42067011_Facial_Recognition_Technology
13. Hashim NMZ, Ibrahim NA, Saad NM, Sakaguchi FZ.
8. References Zakaria International Journal of Emerging Trends &
1. Derya Akkaynak Tali Treibitz University of Haifa Sea- Technology in Computer Science (IJETTCS) Web Site:
thru: A Method For Removing Water From Underwater www.ijettcs.org. 2013 July-August, 2(4). ISSN 2278-
Images. 6856. Barcode Recognition System.
2. Mathias Lechner, Ramin Hasani, Alexander Amini, 14. Aamir Khan, Carmelo Mineo, Gordon Dobie, Charles
Thomas A. Henzinger, Daniela Rus and Radu Grosu. Macleod, Gareth Pierce. Journal of Remanufacturing
Neural circuit policies enabling auditable autonomy Vision guided robotic inspection for parts in
3. Thomas Müller, Alex Evans, Christoph Schied, manufacturing and remanufacturing industry, 2021.
Alexander Keller. Instant Neural Graphics Primitives 15. Efimenko M, Ignatev A, Koshechkin K. Review of
with a Multiresolution Hash Encoding. medical image recognition technologies to detect
https://nvlabs.github.io/instant-ngp/ melanomas using neural networks. BMC
4. Mariusz Bojarski, Davide Del Testa, Daniel Bioinformatics. 2020;21:270.
Dworakowski, Bernhard Firner, Beat Flepp, Prasoon https://doi.org/10.1186/s12859-020-03615-1.
Goyal, et al. End to End Learning for Self-Driving 16. Velusamy, Priyadarshini, Karandharaj, Porkumaran.
Cars. Medical image processing schemes for cancer
5. Fabian Stark, Caner Hazırba S, Rudolph Triebel, Daniel detection: A survey, 2014, 1-6.
Cremers. Technical University of Munich, Germany. 10.1109/ICGCCEE.2014.6922267.
CAPTCHA Recognition with Active Deep 17. Jiahai Chen, You Li, Jianping Gong, Yakun Wu.
Learning.https://www.researchgate.net/publication/301 Application of da Vinci surgical robotic system in
620459_CAPTCHA_Recognition_with_Active_Deep_ hepatobiliary surgery. International Journal of Surgery
Learning and Medicine, 2017, 4(1). 10.5455/ijsm.da-Vinci-
6. Erroll Wood, Tadas Baltrušaitis, Charlie Hewitt, surgical-robotic-system-in-hepatobiliary-surgery.
Sebastian Dziadzio, Matthew Johnson, 18. Kussia Bulbula, Zhang Fengli, Oluwasanmi Ariyo,
Virginia Estellers, Thomas J Cashman and Owusu Forster, Assefa Maregu, Amenu, Temesgen.
Jamie Shotton. Face analysis in the wild using synthetic Vehicle Accident and Traffic Classification Using Deep
data alone. microsoft.github.io/FaceSynthetics/ Convolutional Neural Networks, 2020.
7. Nurul Azma Abdullah, Md Jamri Saidi, Nurul Hidayah 19. Ramasamy Balasubramani. A Review on Vehicle
Ab Rahman, Chuah Chai Wen, Isredza Rahmi A Speed Detection using Image Processing. 2017;l(4):23-
Hamidd. Information Security Interest Group (ISIG). 28.
Fakulti Sains Komputer dan Teknologi Maklumat, 20. Sparavigna, Amelia Carolina. Image Processing:
Universiti Tun Hussein Onn Malaysia, 86400 Batu Methods, Applications and Challenges, 2012.
Pahat, Johor, Malaysia. Face Recognition for Criminal 21. Mohanty SP, Czakon J, Kaczmarek KA, Pyskir A,
Identification: And implementation of principal Tarasiewicz P, Kunwar S, et al. Deep Learning for
component analysis for face recognition. Understanding Satellite Imagery: An Experimental
8. Krishna Kumar Patel, Kar A, Jha SN, Khan MA. Survey. Front. Artif. Intell. 2020;3:534696.
Machine vision system: a tool for quality inspection of DOI:10.3389/frai.2020.534696
food and agricultural products. 22. Govindhasamy G Ayappan. Prediction of rainfall using

~ 55 ~
International Journal of Electronics and Microcircuits www.microcircuitsjournal.com

image processing, 2010. 34. Gillala Rekha V, Krishna Reddy, Amit Kumar Tyagi.
23. Canny John. A Computational Approach To Edge CIRUS - Critical Instances removal based Under-
Detection. Pattern Analysis and Machine Intelligence, Sampling - A solution for Class Imbalance, IJHIS.
IEEE Transactions on. PAMI-8, 1986, 679-698. 2020;16(2):55-66.
10.1109/TPAMI.1986.4767851. 35. Kumari S, Vani V, Malik S, Tyagi AK, Reddy S.
24. Petrović Vladimir, Popović-Bozovic Jelena. Towards Analysis of Text Mining Tools in Disease Prediction.
Real-Time Blob Detection in Large Images with In: Abraham A., Hanne T., Castillo O., Gandhi N.,
Reduced Memory Cost, 2016. Nogueira Rios T., Hong TP. (eds) Hybrid Intelligent
25. Sekar K, Tyagi AK. Study of Data Behaviour and Systems. HIS 2020. Advances in Intelligent Systems
Methods for Data Prediction and Analysis, 6th and Computing, vol 1375. Springer, Cham, 2021.
International Conference on Intelligent Computing and https://doi.org/10.1007/978-3-030-73050-5_55
Control Systems (ICICCS), 2022, 1-6. 36. Amit Kumar Tyagi, Poonam Chahal. Artificial
DOI:10.1109/ICICCS53718.2022.9788360. Intelligence and Machine Learning Algorithms, Book:
26. Malik S, Tyagi AK, Mahajan S. Architecture, Challenges and Applications for Implementing
Generative Model, and Deep Reinforcement Learning Machine Learning in Computer Vision, IGI Global,
for IoT Applications: Deep Learning Perspective. In: 2020. DOI: 10.4018/978-1-7998-0182-5.ch008
Pal S., De D., Buyya R. (eds) Artificial Intelligence- 37. Amit Kumar Tyagi, Rekha G. Challenges of Applying
based Internet of Things Systems. Internet of Things Deep Learning in Real-World Applications”, Book:
(Technology, Communications and Computing). Challenges and Applications for Implementing
Springer, Cham, 2022. https://doi.org/10.1007/978-3- Machine Learning in Computer Vision, IGI Global,
030-87059-1_9 2020, 92-118. DOI: 10.4018/978-1-7998-0182-5.ch004
27. Varsha R, Nair SM, Tyagi AK, Aswathy SU, Radha 38. Gillala Rekha, Krishna Reddy V, Amit Kumar Tyagi.
Krishnan R. The Future with Advanced Analytics: A An Earth mover's distance-based undersampling
Sequential Analysis of the Disruptive Technology’s approach for handling class-imbalanced data,
Scope. In: Abraham A., Hanne T., Castillo O., Gandhi International Journal of Intelligent Information and
N., Nogueira Rios T., Hong TP. (eds) Hybrid Intelligent Database Systems, 2020 13(2/3/4).
Systems. HIS 2020. Advances in Intelligent Systems 39. Akshara Pramod, Harsh Sankar Naicker, Amit Kumar
and Computing, 2021, 1375. Springer, Cham. Tyagi. Machine Learning and Deep Learning: Open
https://doi.org/10.1007/978-3-030-73050-5_56 Issues and Future Research Directions for Next Ten
28. Goyal D, Goyal R, Rekha G, Malik S, Tyagi AK. Years, Book: Computational Analysis and
Emerging Trends and Challenges in Data Science and Understanding of Deep Learning for Medical Care:
Big Data Analytics, 2020 International Conference on Principles, Methods, and Applications, 2020, Wiley
Emerging Trends in Information Technology and Scrivener, 2020.
Engineering (ic-ETITE), 2020, 1-8. DOI: 10.1109/ic- 40. Tyagi, Amit Kumar, Rekha G. Machine Learning with
ETITE47903.2020.316. Big Data (March 20, 2019). Proceedings of
29. Malik S, Mire A, Tyagi AK, Arora V. A Novel Feature International Conference on Sustainable Computing in
Extractor Based on the Modified Approach of Science, Technology and Management (SUSCOM),
Histogram of Oriented Gradient. In: Gervasi O. et al. Amity University Rajasthan, Jaipur - India, 2019 Feb
(eds) Computational Science and Its Applications – 26-28.
ICCSA 2020. ICCSA 2020. Lecture Notes in Computer 41. Gillala Rekha, Amit Kumar Tyagi, Krishna Reddy V. A
Science, Springer, Cham, 2020, 12254. Wide Scale Classification of Class Imbalance Problem
https://doi.org/10.1007/978-3-030-58817-5_54 and its Solutions: A Systematic Literature Review,
30. Gillala Rekha V. Krishna Reddy, and Amit Kumar Journal of Computer Science. 2019;15(7):886-929.
Tyagi, “KDOS - Kernel Density based Over Sampling - ISSN Print: 1549-3636.
A Solution to Skewed Class Distribution, Journal of 42. Deekshetha HR, Shreyas Madhav AV, Tyagi AK.
Information Assurance and Security (JIAS). Traffic Prediction Using Machine Learning. In: Suma,
2020;15(2):44-52. 9p. V, Fernando X, Du KL, Wang H. (eds) Evolutionary
31. Gillala Rekha, Amit Kumar Tyagi, Krishna Reddy V. Computing and Mobile Sustainable Networks. Lecture
Solving Class Imbalance Problem Using Bagging, Notes on Data Engineering and Communications
Boosting Techniques, with and without Noise Filter Technologies, Springer, Singapore, 2022, 116.
Method, International Journal of Hybrid Intelligent https://doi.org/10.1007/978-981-16-9605-3_68
Systems. 2019;15(2):67-76.
32. Gillala Rekha V, Krishna Reddy, Amit Kumar Tyagi. A
Novel Approach for Solving Skewed Classification
Problem using Cluster Based Ensemble Approach,
Mathematical Foundations of Computing, 2020
Feb;3(1):1-9.
33. Mishra S, Tyagi AK. The Role of Machine Learning
Techniques in Internet of Things-Based Cloud
Applications. In: Pal S., De D., Buyya R. (eds)
Artificial Intelligence-based Internet of Things
Systems. Internet of Things (Technology,
Communications and Computing). Springer, Cham,
2022. https://doi.org/10.1007/978-3-030-87059-1_4

~ 56 ~