TECHINICAL SEMINAR

Topic: Computer vision

Name : N.SAI KRISHNA

Roll number : 21BT1A6617

Branch : CSE(AIML)

Year & Sem : IV-1

Academic year : 2024 - 2025

Table of Contents Page no

1.Introduction of computer vision

2.Background study

3.Machine learning in computer vision

4.Results and Discussion

5.History of computer vision

6.Types of computer vision

• Semanatic segmentation
• Localization
• Object detection
• Instance segmentation

7.Application of computer vision

8.Challenges of computer vision

9.Future of computer vision

10. Conclusion

11.Reference
1. Introduction
The machine learning and computer vision hopes to bring into the computers the human capabilities for
data sensing, data understanding, and action taking based on the past and present outcomes. The machine
learning and computer vision research is still evolving [1]. Computer vision is an essential part of Internet
of Things, Industrial Internet of Things, and brain human interfaces. The complex human activities are
recognized and monitored in multimedia streams using machine learning and computer vison. There are
numbers of well-established methods for prediction and analysis such as supervised learning, un-
supervised learning, and semi supervised learning. These methods uses the machine learning algorithms
such as support vector machine, KNN etc. The machine learning solutions revolves around data
gathering, training a model, and use the trained model to make predictions. There are models and services
provided by private companies for speech recognition, text analysis, and image classification. One can
use their models through application programming interfaces (API). For instance, Amazon Recognition,
Polly, Lex, Microsoft Azure Cognitive Services, IBM Watson. Object detection and analysis is an
important part of everyday life. Object detection has applications in avoiding traffic collisions, facial
expression recognition and emotional recognition based on human postures. In [2], developed an
automated system to detect the information contained in human faces from images and video with the
help of orientations. The TensorFlow and OpenPose are the software library used in object detection and
computer vision. The traffic detection models uses convolutional neural network, recurrent neural
network (RNN), long short-term memory
(LSTM), gated recurrent unit (GRU), and Bayesian networks. In an intelligent environments sensors
capture the data which is later used for analysis and forecasting [3, 4]. The feature extraction is one of the
task convolution neural network (CNN) accomplishes without information loss for successful object
detection [5]. The supervised learning
of a deep convolutional neural network recognizes faces with a large set of face images [6]. The only
challenge in computer vision and machine learning application is the data annotation/ labelling [7]. The
machine learning algorithms are now running on cloud as a “machine learning as a service”, “cloud
machine learning” [6]. Moreover, companies, such as Amazon, Microsoft, and Google, have machine
learning as a cloud service. The objective of this research study is to investigate and analytically evaluate
applications of machine learning in computer vision. The database searched includes Google scholar
applying advanced search techniques with respect to keywords- “machine learning”, “computer vision”,
“Deep learning”, and “Artificial intelligence”. The initial search resulted in 258 articles, which included
both patent and citation. After examining articles contents and excluding the citations, the number came
down to 175 articles. Finally, 20 articles formed the core part of this research study. There are five
sections. Section 2 corresponds to background study. Section 3 clusters the existing the machine learning
applications in groups. Section 4 presents results and discussions. The last section concludes with
comments and future work

2. Background study
The computer vision and machine learning are two important areas of recent research. The computer
vision computer uses the image and pattern mappings in order to find solutions [8]. It considers an image
as an array of pixels. The computer vision automates the monitoring, inspection, and surveillance tasks
[6]. Machine learning is the subset of artificial intelligence. The automatic analysis/annotation of videos is
the outcome of computer vision
and machine learning. Figure 1 shows the classification, object detection, and instance segmentation.
Figure 2 shows the object detection in images using Tensor flow and Faster-RCNN-Inception-V2 model
in Anaconda environment.

There are three approaches to machine learning and computer vision: supervised, unsupervised, and semi-
supervised learning. The supervised learning has labelled training data. The labelling of data is expensive,
time consuming and requires expertise. On the other hand the semi-supervised learning has some of the
data labelled and others not. The Bayesian network classifiers has the edge for learning with unlabelled
data. Nevertheless, the real problems falls in the unsupervised learning category where patterns evolves
based on clustering.

The machine learning paradigms for computer vision are support vector machines, neural networks, and
probabilistic graphical models. Support vector machines (SVMs) is a subdomain of supervised machine
learning methods and popular in classification [10]. Neural network consists of layered networks of
interconnected processing nodes [11]. Convolutional neural networks (CNNs) is a category of neural
networks used in image recognition and classification. It has neurons with dimensions: width, height and
depth [10]. CNN has gained
popularity recent times due to largely accessible datasets, GPUs, and regularization techniques [10].
OpenCV is a library, which can be integrated with programming languages such as Android, .NET, Java,
iOS on platforms such as Eclipse and Visual Studio in Windows, iOS, and Linux for image processing
and analysis. It is used in image processing, video analysis, object detection, and machine learning. Figure
3 shows the object detection process in
the machine learning and computer vision environment.

Fig 3. Image processing and object detection process

3. Machine learning in Computer vision

The study explored numerous applications of machine learning in computer vision. For example,
segmentation, feature extraction, refining visual models, pattern matching, shape representation, surface
reconstruction, and 4 Author name / Procedia Computer Science 00 (2019) 000–000
Machine learning in computer vision is used in interpreting data contained in car and pedestrian detection
images [12], automatic classification of faults in railway ties using images [10], interpretation of remote
sensing data for geographical information systems [13], differentiating mango verities based on size
attributes [14], extraction of graphical and textual information from document images [15]. Similarly,
other
applications includes gesture and face recognition [16], machine vision [17], recognizing handwritten
characters and digits [18], advanced driver assistance systems [19], behavioural studies [20], and
estimation of human full-body kinematics for a cyclist and pose estimation [11]. Detecting Curb Ramps
in Google Street View such as automatically identification and examination of curb ramps in images [21].
In [22], studied use of computer vision and machine learning in medical science such as cardiovascular
imaging, retinal blood vessels, nuclear medicine, endoscopy, thermography, angiography, magnetic
resonance, ultrasound and microscopy. Machine learning and computer vision has innovative applications
in engineering, medicines, agriculture, astronomy, sports, education etc.
These applications are categories and clustered in Table 1.
4. Results and discussion
In the world of internet, tons of graphical information and images moves but unlike the textual data the
capability to classify and store them according to the special characteristics is a labour intensive task. The
indexing and storing of graphical data requires computer interventions with advanced model based vision
capabilities and learnability.
This study highlights the research in machine learning and computer vision in various domains. The
machine learning and computer vision techniques has reduced the cost, effort, and time in engineering,
science, and technology. An automated system based on machine learning and computer vision detects
the human emotions (likes and dislikes, confidence levels). The probabilistic models predicts human
activities through labelling and pattern recognition. The machine learning and computer vision in
professional sports measures and analyses the
performances of team and individual players. Moreover, it has been in use in industries for predictive
maintenance. The machine and tools replacements in industries on time before failures have significant
impact on the effectiveness and efficiency of the manufacturing units. The public camera and smart
devices with sensors are huge source of data. The computer vision and machine learning techniques when
applied on these data helps in prediction and monitoring the traffics in cities. Figure 4 shows the evolving
research areas in machine learning and computer
vision. This study has found advancing areas of research in this field as- the biological science (19%) and
human activity (19%) followed by traffic management (13%) and professional sports (13%).
 Fig 4. Machine learning and computer research areas

5.History of computer vision

In the late 1960s, computer vision began at universities that were pioneering artificial intelligence. It was
meant to mimic the human visual system as a stepping stone to endowing robots with intelligent
behavior.[11] In 1966, it was believed that this could be achieved through an undergraduate summer
project,[12] by attaching a camera to a computer and having it "describe what it saw".[13][14]

What distinguished computer vision from the prevalent field of digital image processing at that time was a
desire to extract three-dimensional structure from images with the goal of achieving full scene
understanding. Studies in the 1970s formed the early foundations for many of the computer vision
algorithms that exist today, including extraction of edges from images, labeling of lines, non-polyhedral
and polyhedral modeling, representation of objects as interconnections of smaller structures, optical flow,
and motion estimation.[11]

The next decade saw studies based on more rigorous mathematical analysis and quantitative aspects of
computer vision. These include the concept of scale-space, the inference of shape from various cues such
as shading, texture and focus, and contour models known as snakes. Researchers also realized that many
of these mathematical concepts could be treated within the same optimization framework as
regularization and Markov random fields.[15] By the 1990s, some of the previous research topics became
more active than others. Research in projective 3-D reconstructions led to better understanding of camera
calibration. With the advent of optimization methods for camera calibration, it was realized that a lot of
the ideas were already explored in bundle adjustment theory from the field of photogrammetry. This led
to methods for sparse 3-D reconstructions of scenes from multiple images. Progress was made on the
dense stereo correspondence problem and further multi-view stereo techniques. At the same time,
variations of graph cut were used to solve image segmentation. This decade also marked the first time
statistical learning techniques were used in practice to recognize faces in images (see Eigenface). Toward
the end of the 1990s, a significant change came about with the increased interaction between the fields of
computer graphics and computer vision. This included image-based rendering, image morphing, view
interpolation, panoramic image stitching and early light-field rendering.

6. Types of computer vision models

Computer vision is a field of artificial intelligence (AI) that uses machine learning and neural networks to
teach computers and systems to derive meaningful information from digital images, videos and other
visual inputs—and to make recommendations or take actions when they see defects or issues.

1.Semanatic segmentation

• Assigns a class label to pixels using deep learning algorithms.

• Labels each pixel in an image with a corresponding class label.
• Helps computers understand visual information.
• Used for tasks that require absolute precision.
• Produces a dense pixel-wise segmentation map of an image

2.localization
Object localization is an important CV task. It identifies and correctly localizes certain objects within digital
images or videos. Object localization’s main goal is to precisely determine the position of objects of interest
within an image. Upon that, it represents the object with a bounding box.

The first step in object localization is the object detection. Researchers apply a deep learning model to
identify potential objects within an image. The detection step utilizes region proposal networks to identify
and mark regions that probably contain objects.

Upon object detection, precise localization refines the detected regions. It draws bounding boxes that
contain the identified objects. Also, advanced techniques such as instance segmentation outline the
boundaries of objects at the pixel level.

3.object detection
• Uses neural networks to localize and classify objects in images.
•
Identifies and locates objects within an image or video, drawing bounding boxes around them
•
Helps detect instances of visual objects of certain classes in digital images
•
Aims to identify and annotate each object present in the media
 • Involves image localization using bounding boxes

4.Instance segmentation

Instance segmentation is a type of image segmentation that involves detecting and segmenting each
object in an image. It is similar to object detection but with the added task of segmenting the object’s
boundaries. The algorithm has no idea of the class of the region, but it separates overlapping objects.
Instance segmentation is useful in applications where individual objects need to be identified and tracked.

7.Aplications of computer vision

Computer vision or machine vision is a field of science that enables computers or devices to recognize
different objects just like human beings. the computers need to be trained to detect objects and also some
patterns just like you teach a kid to identify the objects but the computers are more efficient as it takes
very little time to be trained. Computer vision has applications in all industries and sectors and they are
as follows:

1. Oil and natural gas: The oil and natural gas companies produce millions of barrels of oil and billions
of cubic feet of gas every day but for this to happen, first, the geologists have to find a feasible location
from where oil and gas can be extracted. To find these locations they have to analyze thousands of
different locations using images taken on the spot. Suppose if geologists had to analyze each image
manually how long will it take to find the best location? Maybe months or even a year but due to the
introduction of computer vision the period of analyzing can be brought down to a few days or even a few
hours. You just need to feed in the images taken to the pre-trained model and it will get the work done.

2.Hiring process: In the HR world, computer vision is changing how candidates get hired in the interview
process. By using computer vision, machine learning, and data science, they’re able to quantify soft skills
and conduct early candidate assessments to help large companies shortlist the candidates.

3.Video surveillance: The Concept of video tagging is used to tag videos with keywords based on the
objects that appear in each scene. Now imagine being that security company who’s asking to look for a
suspect in a blue van amongst hours and hours of footage. You will just have to feed the video to the
algorithm. With computer vision and object recognition, searching through videos has become a thing of
the past.

4. Construction: Take for example the electric towers or buildings, which require some degree of
maintenance to check for degrees of rust and other structural defects. Certainly, manually climbing up the
tower to look at every inch and corner would be extremely time-consuming, costly, and dangerous. Flying
a drone with wires around the electric tower doesn’t sound particularly safe either. So how could you apply
computer vision here? Imagine that if a person on the ground took high-resolution images from different
angles. Then the computer vision specialist could create a custom classifier and use it to detect the flaws
and amount of rust or cracks present.
5. Healthcare: From the past few years, the healthcare industry has adopted many next-generation
technologies that include artificial intelligence and machine learning concept. One of them is computer
vision which helps determine or diagnose disease in humans or any living creatures.

6.Agriculture: The agricultural farms have started using computer vision technologies in various forms
such as smart tractors, smart farming equipment, and drones to help monitor and maintain the fields
efficiently and easily. It also helps improve yield and the quality of crops.

7.Military: For modern armies, Computer Vision is an important technology that helps them to detect
enemy troops and it also enhances the targeting capabilities of guided missile systems. It uses image sensors
to deliver battlefield intelligence used for tactical decision-making. One more important Computer Vision
application in the areas of autonomous vehicles like UAV’s and remote-controlled semi-automatic vehicles,
which need to navigate challenging terrain

8. Industry: In manufacturing or assembly line, computer vision is being used for automated inspections,
identifying defective products on the production line, and for remote inspections of machinery. The
technology is also used to increase the efficiency of the production line.

9. Automotive: This is one of the best examples of computer vision technologies, which is a dream come
true for humans. Self-driving AI analyzes data from a camera mounted on the vehicle to automate lane
finding, detect obstacles, and recognize traffic signs and signals.

10. Automated Lip Reading: This is one of the practical implementations of computer vision to help
people with disabilities or who cannot speak, it reads the movement of lips and compares it to already
known movements that were recorded and used to create the model.

8.Challenges of computer vision

• 1. High costs Comparing computer vision platform costs isn’t easy, as they have high variability
when it comes to functionality. ...
• 2. Lack of experienced professionals With more companies developing and implementing
computer vision applications, it’s important to have the right professionals to bring ideas to
reality. ...
• 3. Size of required data sets ...
• 4. Need for regular monitoring ...
1. Growing demand for computer vision solutions

With the global computer vision market expected to reach a value of US$30.94 billion by 2028, it’s no
wonder that there’s been an increase in the demand for its solutions.

Businesses and consumers are adopting computer vision applications like object recognition, vehicle
tracking, and face recognition. As the market expands, so does the need for skilled workers capable of
developing and maintaining those applications.

Companies like Intel are investing in the autonomous vehicle industry, which – expected to rise to US$7
trillion by 2050. Netflix, for example, uses computer vision to harvest static image frames for digital
merchandising. With one single episode of Stranger Things having almost 86,000 static video frames, it’s
easy to see why.

2. Increasing need for automation

The demand for workflow automation processes alone is expected to rise to US$26 billion by 2025.
Automation lets machines complete tasks with minimal to no human intervention, which frees employees
from repetitive actions that machines can perform more efficiently.

Artificial intelligence and automation are changing how businesses operate, with new generations of robots
appearing in areas like automated check-outs in grocery stores. SpaceX, for example, has used AI and
automation to help astronauts dock successfully into the International Space Station (ISS).

3. Increasing demand for vision-guided robotic systems

Vision-guided robotic systems are machines that use cameras and sensors to better understand their
environment. This market is divided into 2D and 3D vision-guided robotics. 2D robotic systems don’t have
depth perception, and their accuracy is limited when working with precision tasks. They’re still useful,
however, in both industrial and non-industrial settings. 3D robotic systems have a single camera lens for
accurate images of their surroundings. They can produce dynamic images for better accuracy and can work
in both static and dynamic environments.
4. Growing adoption of Industry 4.0

Governments stimulate manufacturing by investing in research and development programs. This allows for
innovative technology research and higher economic growth.

The automated industrial market is expected to grow exponentially, as the demand for industrial automation
rises. Behind this adoption are government incentives, rising manual labor costs, and strict workplace
regulations, as well as more SMEs adopting automation.

5. Growing demand for AI in computer vision

The market for artificial intelligence in computer vision is expected to be worth US$144.46 billion by 2028.
The rising demand for automation, quality inspection, and computer vision systems is behind this growth.

Computer vision systems and software are continuously being developed for new applications in industries
such as traffic management, healthcare, document processing, security and surveillance, and more.

AI-enabled computer vision systems are increasingly popular, with potential in areas like consumer
electronics.

6. Increased manufacturing of autonomous vehicles

The autonomous vehicle market is predicted to rise in value to US$11.03 billion by 2028. These self-driving
cars can read and sense their surroundings, operating with little- to- no human input.

There has been a rapid evolution in high-definition mapping, sensor-processing technologies, adaptive
algorithms, and deployment of infrastructure-to-vehicle and vehicle-to-vehicle communication technologie

4 top challenges in computer vision

The size of required data sets was highlighted as the biggest limitation in computer vision, receiving 42.3%
of votes. High costs (at 20.4%) were indicated as the second biggest limitation.It’s also worth noting that
those who responded with “Other” mentioned limitations including a lack of suitable silicon, privacy or
data security, lack of data, and lack of resources Now, let’s take a closer look at the biggest limitations of
computer vision:

1. High costs

Comparing computer vision platform costs isn’t easy, as they have high variability when it comes to
functionality. It also leads to unexpected costs when software products do not provide the necessary
infrastructure, or when they need additional products.An application’s architectural design and
infrastructure are big determinants of total costs. A web application that analyzes a few images per minute
has vastly different requirements than an always-on computer vision system that processes multiple video
streams in real-time.As computer vision involves resource-intensive complex tasks like image
recognition, powerful processors and complex software increase expenses exponentially. But computer
vision doesn’t have to be expensive:

Focus on privacy and security. Breaches of privacy or security can lead to extremely high costs, even
though implementing best security practices is also expensive. It’s vital to have computer vision running
on monitored and hardened infrastructure, with all data and communication encrypted.

Use image pre-processing. Pre-processing methods optimize and standardize images before they are fed
into a deep learning model, giving the model better accuracy.

Run computer vision at the edge. It can be more cost-effective to run computer vision at the edge,
especially for scalable, low-latency applications, in comparison to Cloud APIs.

Use low-code/no-code engineering. Reducing the amount of manual coding speeds up application
delivery, allowing for building, deploying, and updating at much faster speeds with reduced risk.

Use next-gen deep learning models. Up-to-date frameworks make big leaps in real-time in techniques like
real-time object detection performance.

Invest in cross-platform solutions. Depending on just one machine learning platform or hardware
manufacturer platform isn’t ideal. To decrease costs, it’s important to have the ability to use and exchange
cross-platform hardware and software.

2. Lack of experienced professionals

With more companies developing and implementing computer vision applications, it’s important to have
the right professionals to bring ideas to reality. With hundreds of thousands of artificial intelligence
professionals across the globe but millions of roles available, the skills gap needs to be bridged.

The demand for professionals has increased with the democratization of artificial intelligence and deep
learning applications, which led to corporations and startups alike beginning work in the field.
The skills gap can be addressed by increasing resources for digital, technical, and math education.
Companies like Amazon and Google invest internationally to expand their talent pool, but companies can
also focus on retraining and upskilling current employees.

3. Size of required data sets

Models are only as good as their data. Models should train on more examples than trainable parameters.
Google, for example, trains simple linear regression models on large data sets. A small data set, with 150
examples, won’t be as successful as one with trillions of examples.

But quantity isn’t everything – quality matters too. By choosing the option that offers the best outcome, a
quality data set will accomplish its task. When collecting data, having a more defined idea of quality is
important. Better-performing models have:

Reliability. How much can you trust the data? Are there a lot of label errors? Are features noisy? Is data
properly filtered?

Feature representation. The mapping of data to useful features – how is the data shown on the model?
Should numeric values be normalized? How should outliers be handled?

Minimized skew. Different results can be computed for metrics at training time versus serving time. Skew
can be subtle but completely destroy your results, so keep in mind what data is available to your model at
prediction time.

4. Need for regular monitoring

Technical issues and breakdowns to computer vision systems can prove incredibly disruptive to
businesses. Companies need a dedicated in-house team to monitor and evaluate their computer vision
systems.

This team needs to conduct regular and unplanned maintenance, and update and change algorithms as
required. Testing changes can be a long and time-intensive process.

9.Future of computer vision

Computer vision is a rapidly growing field in research and applications. Advances in computer vision

research are now more directly and immediately applicable to the commercial world.

AI developers are implementing computer vision solutions that identify and classify objects and even

react to them in real time. Image classification, face detection, pose estimation, and optical flow are some

of the typical tasks. Computer vision engineers are a subset of deep learning (DL) or machine learning

(ML) engineers that program computer vision algorithms to accomplish these tasks.
The structure of DL algorithms lend themselves well to solving computer vision problems. The

architectural characteristics of convolutional neural networks (CNNs) enable the detection and extraction

of spatial patterns and features present in visual data.

The field of computer vision is rapidly transforming industries like automotive, healthcare, and robotics,

and it can be difficult to stay up-to-date on the latest discoveries, trends, and advancements. This post

highlights the core technologies that are influencing and will continue to shape the future of computer

vision development in 2022 and beyond:

• Cloud computing services that help scale DL solutions.

• Automated ML (AutoML) solutions that reduce the repetitive work required in a standard ML
pipeline.
• Transformer architectures developed by researchers that optimize computer vision tasks.
• Mobile devices incorporating computer vision technology.

Cloud computing

Cloud computing provides data storage, application servers, networks, and other computer system

infrastructure to individuals or businesses over the internet. Cloud computing solutions offer quick, cost-

effective, and scalable on-demand resources.

Storage and high processing power are required for most ML solutions. The early-phase development of

dataset management (aggregation, cleaning, and wrangling) often requires cloud computing resources for

storage or access to solution applications like BigQuery, Hadoop, or BigTable.

Figure 1. Interconnected data center, representing the need for cloud computing and cloud services
(Photo by Taylor Vick on Unsplash)

Recently, there has been a notable increase in devices and systems enabled with computer vision

capabilities, such as pose estimation for gait analysis, face recognition for smartphones, and lane detection

in autonomous vehicles.

The demand for cloud storage is growing rapidly, and it is projected that this industry will be valued at

$390.33B—five times the market’s current value in 2021. The increased market size will lead to an increase

in the use of inbound data to train ML models. This correlates directly to larger data storage capacity

requirements and increasingly more powerful compute resources.

GPU availability has accelerated computer vision solutions. However, GPUs alone aren’t always enough

to provide the scalability and uptime required by these applications, especially when servicing thousands

or even millions of consumers. Cloud computing provides the needed resources to startup and supplement

existing on-premises infrastructure gaps.

Cloud computing platforms, including Amazon Web Services (AWS), Google Cloud Platform (GCP), and

Microsoft Azure, provide end-to-end solutions to core components of the ML and data science project

pipeline, including data aggregation, model implementation, deployment, and monitoring. For computer
vision developers designing vision systems, it’s important to be aware of these major cloud service

providers their strengths, and how they can be configured to meet specific and complex pipeline needs.

Computer vision at scale requires cloud service integration

The following are examples of NVIDIA services that support typical computer vision systems.

The NGC Catalog of pretrained DL models reduces the complexity of model training and implementation.

DL scripts provide ready-made customizable pipelines. The robust model deployment solution automates

delivery to end users.

NVIDIA Triton Inference Server enables the deployment of models from frameworks such as TensorFlow

and PyTorch on any GPU– or CPU-based infrastructure. Triton Inference Server provides scalability of

models across various platforms, including cloud, edge, and embedded devices.

The NVIDIA partnership with cloud service providers such as AWS enables the deployment of computer

vision-based assets, so computer vision engineers can focus more on model performance and optimization.

Businesses reduce costs and optimize strategies wherever feasible. Cloud computing and cloud service

providers accomplish both by providing billed solutions based on usage and scaling based on demand.

AutoML

ML algorithms and model development involve a number of tasks that can benefit from automation like,

feature engineering and model selection.

Feature engineering involves the detection and selection of relevant characteristics, properties, and

attributes from datasets.

Model selection involves evaluating the performance of a group of ML classifiers, algorithms, or solutions

to a given problem.
Both feature engineering and model selection activities require considerable time for ML engineers and

data scientists to complete. Software developers frequently revisit these phases of the workflow to enhance

model performance or accuracy.

Figure 2. AutoML enables the automation of repetitive tasks such as numeric calculations
(Photo by Stephen Dawson on Unsplash)

There are several large ongoing projects to simplify the intricacies of an ML project pipeline. AutoML

focuses on automating and augmentation workflows and their procedures to make ML easy accessible, and

less manually intensive for non-ML experts.

Looking at the market value, projections expect the AutoML market to reach $14 billion by 2030. This

would mean an increase ~42x higher than its current value.

This particular marriage of ML and automation is gaining traction, but there are limitations.
AutoML in practice

AutoML saves data scientists and computer engineers time. AutoML capabilities enable computer vision

developers to dedicate more effort to other phases of the computer vision development pipeline that best

use their skillset like model training, evaluation, and deployment. AutoML helps accelerate data

aggregation, preparation, and hyperparameter optimization, but these parts of the workflow still require

human input.

Data preparation and aggregation are needed to build the right model, but they are repetitive, time-

consuming tasks that depend on locating appropriate data quality sources.

Likewise, hyperparameter tuning can take a lot of time to iterate to get the right algorithm performance. It

involves a trial-and-error process with educated guesses. The amount of repeated work that goes into

finding the appropriate hyperparameters can be tedious but critical for enabling the model’s training to

achieve the desired accuracy.

For those interested in exploring GPU-powered AutoML, the widely used Tree-based Pipeline

Optimization Tool (TPOT) is an automated ML library aimed at optimizing ML processes and pipelines

through the utilization of genetic programming. RAPIDS cuML provides TPOT functionalities accelerated

with GPU compute resources. For more information, see Faster AutoML with TPOT and RAPIDS.

Machine learning libraries and frameworks

ML libraries and frameworks are essential elements in any computer vision developer’s toolkit. Major DL

libraries such as TensorFlow, PyTorch, Keras, and MXNet received continuous updates and fixes in 2021,

and will likely continue to do so in the future.

More recently, there have been exciting advances going on in mobile-focused DL libraries and packages

that optimize commonly used DL libraries.

MediaPipe extended its pose estimation capabilities in 2021 to provide 3D pose estimation through the

BlazePose model, and this solution is available in the browser and on mobile environments. In 2022, expect
to see more pose estimation applications in use cases involving dynamic movement and those that require

robust solutions, such as motion analysis in dance and virtual character motion simulation.

PyTorch Lightning is becoming increasingly popular among researchers and professional ML practitioners

due to its simplicity, abstraction of complex neural network implementation details, and augmentation of

hardware considerations.

State-of-the-art deep learning

DL methods have long been used to tackle computer vision challenges. Neural network architectures for

face detection, lane detection, and pose estimation all use deep consecutive layers of CNNs. A new

architecture for computer vision algorithms is emerging: transformers.

The Transformer is a DL architecture introduced in Attention Is All You Need. The paper methodology

creates a computational representation of data by using the attention mechanism to derive the significance

of one part of the input data relative to other segments of the input data.

The Transformer does not use the conventions of CNNs, but research has shown the applications of

transformer models in vision-related tasks. Transformers have made a considerable impact within the NLP

domain. For more information, see Generative Pre-trained Transformer (GPT) and Bidirectional Encoder

Representation From Transformer (BERT).

Explore a transformer model through the NGC Catalog that includes details of the architecture and

utilization of an actual transformer model in PyTorch.

For more information about applying the Transformer network architecture to computer vision, see the

Transformers in Vision: A Survey paper.

Mobile devices

Edge devices are becoming increasingly powerful. On-device inference capabilities are a must-have feature

for mobile applications used by customers who expect quick service delivery and AI features.
Figure 3. Mobile devices are a direct commercial application of computer vision features
(Photo by Homescreenify on Unsplash)

10.Conclusion
The commercial and academic research on computer vision is growing in the form of new techniques,
process, models, and algorithms. Machine learning has been able to address many issues of feature
extractions and processing in computer vision. The machine learning and computer vision synthesis has
helped in understanding complex problems. The machine learning applications in computer vision have
varied outputs depending upon the domain. This study includes the analysis, classification, and discussion
on the use of machine learning in computer vision. The research has identified the successful
implementation of machine learning applications in computer vision for weather forecasting, biological
science, expression reading, food security, species classification, sports, monitoring the traffic flows, and
predictive maintenance in industries. The biological science, human activity interpretation, traffics
management, and professional sports are the emerging areas. The object detection, classification, and
prediction are the most frequent use of machine learning in computer vision. The future work
would assess accuracy of the machine learning algorithms in computer vision.

Computer Vision, an interdisciplinary field at the intersection of artificial intelligence and image
processing, focuses on enabling machines to interpret and understand visual data from the world around
us. This technology empowers computers to derive meaningful information from images, videos, and
other visual inputs, facilitating various applications across various industries.
computer vision is a powerful and fast-growing technology that is already being used in many areas of
our lives, from autonomous vehicles to facial recognition. It has the potential to benefit humanity in a
variety of ways. As technology continues to advance, so too does the potential of CV and its ability to
create more sophisticated artificial intelligence systems. Therefore, it is important that we continue to
invest in research and development to ensure that the advantages of this technology are maximized.
11.References
[1] Kerle, Norman, Markus Gerke, and Sébastien Lefèvre, GEOBIA 2016: Advances in Object-Based
Image Analysis—Linking with Computer
Vision and Machine Learning. 2019.
[2] Sebe, Nicu, Michael S Lew, Yafei Sun, Ira Cohen, Theo Gevers, and Thomas S Huang, Authentic
facial expression analysis. Image and
Vision Computing, 2007. 25(12): p. 1856-1863.
[3] Kaloskampis, Ioannis, Recognition of complex human activities in multimedia streams using machine
learning and computer vision. 2013,
Cardiff University.
[4] Al-Badi, Ali, Ali Tarhini, and Asharul Islam Khan. Exploring Big Data Governance Frameworks. in
9th International Conference on
Emerging Ubiquitous Systems and Pervasive Networks (EUSPN 2018). 2018. Leuven, Belgium: Procedia
Computer Science.
[5] Zhang, Fan, Wei Li, Yifan Zhang, and Zhiyong Feng, Data Driven Feature Selection for Machine
Learning Algorithms in Computer
Vision. IEEE Internet of Things Journal, 2018. 5(6): p. 4262-4272.
[6] Bond, Raymond, Ansgar Koene, Alan Dix, Jennifer Boger, Maurice D Mulvenna, Mykola Galushka,
Bethany Waterhouse Bradley, Fiona
Browne, Hui Wang, and Alexander Wong, Democratisation of Usable Machine Learning in Computer
Vision. arXiv preprint
arXiv:1902.06804, 2019.
[7] Fathi, Alireza, Recent Hot Machine Learning Hammers used in Computer Vision. 2012.
[8] Esposito, Floriana and Donato Malerba, Machine learning in computer vision. Applied Artificial
Intelligence, 2001. 15(8): p. 693-705.