Leveraging Event Marketing Performance using AI in Facial Recognition
Peter Khallouf
1
and Christine Markarian
2
1
Data Science - Data and IT, International University of Applied Sciences, Germany
2
Department of Engineering and Information Technology, University of Dubai, U.A.E.
Keywords:
Artificial Intelligence, Facial Recognition Algorithm, Machine Learning, Engagement Measurement, Event
Marketing.
Abstract:
With the advances in technology and the rapid changes in human-technology interactions, the event mar-
keting field has seen major developments over the past years. Despite its remarkable growth, many aspects
of event marketing do not yet align with the best available technologies. In this paper, we aim to leverage
event marketing performance using artificial intelligence techniques. We design a framework that optimizes
attendee-feedback generation using a facial-recognition algorithm. The framework measures attendees’ en-
gagement levels by periodically extracting attendee facial features during a session, categorizing them into
seven states of emotions (anger, disgust, fear, happiness, neutral, sadness, and surprise), and then analyzing
session engagements based on the obtained results. These measurements are then used to give insights about
an event’s performance during and after sessions, thus improving the overall performance of a given event.
The proposed framework is easy-to-implement, time-efficient, and cost-effective.
1 INTRODUCTION
Over the past decades, marketing has become one of
the most prominent circulating topics, steering most
of our activities, decision-making, and influences.
It has played a major success role in most indus-
tries. With the advances in technology and the rapid
changes in human-technology interactions, the event
marketing field has seen major developments (Hoyle,
2016; Gupta, 2003). A survey of 700 companies
conducted by Harvard Business Review Analytic Ser-
vices shows that 93 percent of companies put a lot of
effort into hosting events, 57 percent of which con-
sider it a very high priority (Services, 2018). Despite
its remarkable growth, many aspects of event market-
ing do not yet align with the best available technolo-
gies. While a lot of research has been dedicated to
studying the evolving role of artificial intelligence in
marketing in general (Sterne, 2017; De Bruyn et al.,
2020; Huang and Rust, 2021; Huang and Rust, 2021),
only few works have targeted event marketing in par-
ticular (Neuhofer et al., 2020).
In this paper, we focus our study on leveraging
event marketing performance using artificial intelli-
gence techniques.
Event marketing is known as three main goals:
generating brand awareness, measuring engagement
during an event, and educating the audience. Engage-
ment during an event is measured by observing atten-
dees in their sessions attended, meetings held, surveys
conducted, and social media activities. Generating at-
tendee feedback from these observations is often done
using traditional techniques which are mostly prone
to high level of inaccuracy and unreliability. In gen-
eral, people have a tendency, for many reasons, not
to reveal their true opinions about events or products.
They sometimes do not want to put effort into ex-
pressing their likes and dislikes - the reason why many
surveys are filled recklessly and inattentively. Gath-
ering proper feedback would require constantly fol-
lowing up with attendees, which not only incurs ad-
ditional costs, but does not give the correct insights,
since many people would simply not want to par-
ticipate in giving feedback. Such inaccuracies lead
to degradation in the Return On Investment (ROI).
Hence, a slight improvement in the techniques used to
generate customer feedback would imply higher suc-
cess rates in profit maximization.
Motivated by the above, we develop in this pa-
per a novel framework that leverages event marketing
performance by improving attendee feedback gener-
ation using Artificial Intelligence (AI) techniques in
facial recognition. The idea is to observe attendees
during the sessions and identify, based on their fa-
cial expressions, whether they feel engaged or not.
568
Khallouf, P. and Markarian, C.
Leveraging Event Marketing Performance using AI in Facial Recognition.
DOI: 10.5220/0010864800003116
In Proceedings of the 14th International Conference on Agents and Artificial Intelligence (ICAART 2022) - Volume 3, pages 568-573
ISBN: 978-989-758-547-0; ISSN: 2184-433X
Copyright
c
2022 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
This gives companies insights about their presenta-
tions, trainings, and products, without the need to
reach out to potential customers to ask about their
feedback. Moreover, given that facial expressions are
known to be a direct translation of true emotions (Ek-
man, 1993; Niedenthal et al., 2000), the information
gathered in this framework would be way more accu-
rate compared to those gathered using the traditional
approaches.
Artificial Intelligence (AI) techniques have been
widely used in marketing. (Campbell et al., 2020)
studied the revolution of AI in business and society,
aiming to understand and engage customers. (Kietz-
mann et al., 2018) used AI to understand and guide
consumers with the help of machine learning tech-
niques, by collecting data from different sources and
combining them to gain on-the-spot consumer in-
sights. (Whitehill et al., 2014) calculated the relia-
bility of human observers’ judgments to learn about
students’ engagement from their faces.
To the best of our knowledge, there has been no
study about event engagement measurement using AI.
1.1 Our Contribution
In this paper, we propose a new engagement mea-
surement framework that aims to optimize attendee
feedback generation. The framework uses a facial-
recognition algorithm to classify attendees based on
seven states of emotions (anger, disgust, fear, hap-
piness, neutral, sadness, and surprise) and accord-
ingly calculates their average engagement level in a
given session. Our proposed approach is easy-to-
implement, time-efficient, and cost-effective.
At the heart of the framework is a software that de-
ploys a high-definition camera to extract attendee fa-
cial features at specific session times (including time
spent in a session). The software inputs the extracted
pictures into a facial recognition algorithm, resulting
in the attendees’ states of emotions. The records are
saved in a database and are used to calculate the av-
erage session engagements. These engagements are
then used to give insights about the success of a given
presentation or an event in general. Moreover, what
makes the framework even more powerful is the fact
that the results of session engagements obtained from
the software could be displayed instantly to session
presenters. Hence, if during a session, the majority
of the states of emotions express confusion, the ses-
sion presenter can shift immediately to other means of
communication to get back attendees’ attention. The
software would give the proper alerts instantly during
a presentation so as to make the best out of a given
session.
To output attendees’ states of emotions, our
framework uses the facial recognition algorithm pro-
posed in (Richhariya and Gupta, 2019). This algo-
rithm is shown to outperform most facial-recognition
algorithms available so far in the literature, in terms
of speed, accuracy, and implementation cost.
1.2 Outline
The remainder of the paper is structured as follows.
In Section 2, we give an overview of works related to
AI-based event marketing and facial recognition. In
Section 3, we present the algorithmic, hardware, and
software requirements of the proposed framework. In
Section 4, we present the engagement measurement
framework and analyze it in Section 5. In Section 6,
we extend the scope of the framework to event orga-
nizers and attendees. We conclude in Section 7 with
a discussion and future works.
2 RELATED WORK
In this section, we give an overview of AI-based event
marketing and facial recognition studies related to our
work.
2.1 AI-based Event Marketing
AI has been intensively used to enhance marketing
strategies. A study in (Campbell et al., 2020) shows
the revolution of AI in business and society in terms of
predicting, understanding, and engaging customers.
The study is based on using nine marketing functions
in the marketing planning process to consolidate how
AI can enhance it. The nine components are inherent
to strategic marketing, providing an organizing frame-
work to assist marketing professionals with practical
uses of AI.
(Kietzmann et al., 2018) studies the importance
of AI in the way advertisers understand and guide
consumers. With the help of machine learning tech-
niques, advertisers are able to collect consumers’ data
from different sources, combine the collected data,
and mine them, to deliver on-the-spot consumer in-
sights. The study highlights on the role of AI in help-
ing consumer and advertisers at the same time, by giv-
ing insights that take into account the public’s privacy
rights.
2.2 Facial Recognition
Our framework deploys the facial recognition algo-
rithm proposed in (Richhariya and Gupta, 2019). The
Leveraging Event Marketing Performance using AI in Facial Recognition
569
algorithm is based on using the so-called Univer-
sum data to conduct multiclass face emotion catego-
rization from human facial pictures. Unlike previ-
ous Universum-based models which have high train-
ing costs, the proposed algorithm, known as, It-
erative Universum Twin Support Vector Machine
(IUTWSVM), outperforms the previous models. It
is characterized by high accuracy in detecting facial
expression, speed, and low implementation cost.
A study about automatic recognition of student
engagement from facial expressions in (Whitehill
et al., 2014) has been an inspiration for our work in
this paper. The study considers human observers’
judgement, studies the signals made from these judge-
ments, and uses machine learning techniques to auto-
mate the process. A comparison between human ob-
servers and AI-based observation is made. Low-cost
computers with a high-resolution camera are used to
monitor the engagement of students in a given class.
This information is used to understand when and why
students get disengaged and to accordingly shift to
other teaching methods.
3 TECHNICAL REQUIREMENTS
To implement the solution and integrate between the
infrastructure, the high-definition camera, and the
facial-recognition algorithm, a number of hardware
and software requirements are to be met. In this sec-
tion, we give an overview of the facial-recognition al-
gorithm and discuss these requirements.
3.1 Facial-Recognition Algorithm
Facial recognition is generally performed in five
steps (Woodward Jr et al., 2003):
1. An image taken by a camera is acquired.
2. The locations of faces are detected in the acquired
image.
3. Each detected face is analyzed based on the spa-
tial geometry of face features. There are different
ways to extract these features, the most common
method is known as Principle Components Anal-
ysis (PCA). This process yields a so-called tem-
plate for each face, which is a reduced set of data
representing the unique face features.
4. The generated template is compared to templates
of known faces stored in a database. This process
results in scores that indicate how similar the gen-
erated template is in comparison to those in the
database. In case of identity verification, the gen-
erated template is compared to only one template
Figure 1: Findface Camera Interface Detecting Several
Faces (Findface, 2021).
in the database, that associated with the claimed
identity.
5. The last step determines whether the scores ac-
quired are high enough to declare a match.
We refer the reader to (Richhariya and Gupta,
2019) for a detailed description of how the algorithm
performs these steps.
3.2 Hardware Requirements
Unlike other IT solutions that normally require a
server and sometimes with high specifications, our
proposed solution does not deploy a server. Instead,
one computer with the following specifications is
used, to share the resources among the different parts
of the solution. The computer needed would run on
a Windows 7 operating system and have the follow-
ing specifications: 64-bit architecture, 3.20 GHz Intel
Core i7 2400 processor, 8 GB of RAM, Nvidia video
card 3Gb of video functionality. We can additionally
add a redundant computer to cover the high availabil-
ity plan and failover coverage. However, this is only
needed for robustness as a form of a contingency plan
associated with the event’s success.
A high-definition camera is used. We propose the
FindFace camera, which is recognized in many world-
wide projects and is highly reliable in terms of hard-
ware robustness and service support services (Find-
face, 2021). It is easily available and has relatively
low costs. The FindFace camera is able to simulta-
neously identify a significantly large number of faces
and send notifications at a very high speed - it can
detect, identify, and notify about, all the images avail-
able within the same frame in 0.5 seconds. Figure 1
shows the camera’s interface detecting several faces
at a time. Our framework uses this feature to allow
a presenter in an event to shift between presentation
methods based on the notifications received.
ICAART 2022 - 14th International Conference on Agents and Artificial Intelligence
570
3.3 Software Requirements
The computer would host a database that saves all the
inward transactions from the camera using the man-
ufacturer’s Software Development Kit (SDK). These
transactions would contain the attendees’ images that
will be fed to the facial recognition algorithm. The
latter returns the state of emotion (anger, disgust, fear,
happiness, neutral, sadness, and surprise) for each im-
age, which will be saved in the database. At the end
of the event, the database will contain loads of data
that can be expressed in figures and translated into in-
sights. Depending on the investor’s database licensing
budget, any of the following databases could be used:
SQL Server, MySQL, or Oracle.
MATLAB R2008b will be installed on the com-
puter to implement the facial recognition algorithm.
The software can be developed using any .Net lan-
guage. Its functionalities are straightforward - it com-
municates with the camera, sends the images to the
facial recognition algorithm, saves the transactions in
the database, and reports tools to visualize the final
results. We give a detailed software architecture in
Section 4.
4 ENGAGEMENT
MEASUREMENT
FRAMEWORK BASED ON
FACIAL RECOGNITION
The engagement measurement framework builds
upon the following steps.
1. The FindFace camera is installed in the exact lo-
cation (kiosk) where a given presentation is taking
place.
2. The FindFace camera will be connected to the
main network where the computer is deployed.
3. A number of software will be installed to the com-
puter, namely, the Database Management System
(DBMS), MATLAB R2008b, and the FindFace
driver.
4. A database with the required tables will be cre-
ated to hold the incoming information from the
camera and the facial-recognition algorithm. This
database will be installed to the Database Man-
agement System (DBMS).
5. The DBMS’s high availability will be configured
to the other computer to prevent any possible
failover. As described in Section 3, this will only
be added in case a robust solution is desirable.
Figure 2: Flowchart/ Process Diagram.
Figure 3: Software Design Structure.
6. The below software will be developed. Its job
will be to communicate with the different hard-
ware and software components of the solution.
The software will be developed using any .Net
language, preferably C-Sharp. The latter is a high-
level language and is easy to use. The implemented
software will be straightforward and hence easy to im-
plement. Its functionality is described as follows - we
refer the reader to Figure 2 for an illustration.
integrate, via the SDK, with the FindFace camera
receive the captured images
send the received images to MATLAB for feature
extraction
save the results from MATLAB in the database
generate a report based on the treated data in the
database
Next, we describe the software’s design in terms
of its view, data, and logic layers - we refer the reader
to Figure 3 for an illustration.
View Layer: The software’s view layer includes
two tabs. The first tab contains the FindFace cam-
era and the database controls to input the connectivity
link. Other controls display the connectivity status
and the number of images saved in the database, after
Leveraging Event Marketing Performance using AI in Facial Recognition
571
being processed by the facial recognition algorithm.
The database will contain a record for each attendee
with information about the time stamp, presentation
session, and each emotional status during the presen-
tation. These controls are needed to monitor and en-
sure that the system functions properly. The second
tab contains the dashboard that will display the results
associated with the event marketing evaluation in the
form of graphs and spreadsheets. These results will
contain the live data as well as the conclusive data
at the end of the event, of a given presentation. At
this point, one can measure the engagement level in
the presentation based on the insights concluded from
these data.
Data Layer: The data layer uses transactions in all
the database classes to ensure the transaction’s roll-
out during interruptions. We create the database con-
nection class, referenced in a separate class that will
contain the Data Manipulation Language (DML).
Logic Layer: The logic layer is coded in a scalable
way using classes and objects to ensure its adaptive-
ness with future solution enhancements and different
integrations. The facial recognition algorithm will
have a separate class, which will return the state of
emotion for the corresponding received image. The
main form will run a thread having the ordered func-
tion as follows:
Create a record in the database for each attendee
once the presentation starts.
Run the facial recognition algorithm periodically,
such as, every one minute, and update the results
associated with all attendees in the database.
Update a flag in the database, during the presen-
tation, about the session’s status (in progress, fin-
ished)
The presenter’s computer will be linked to the
same network as the software, and the software
will integrate with the presentation tools to record
the start/end of the presentation.
5 PERFORMANCE
Unlike traditional feedback generation approaches
that require extensive resources in terms of time, cost,
and processing, the proposed framework is able to
output large amounts of insightful information within
milliseconds.
The framework implements a facial-recognition
algorithm, rather than relying on the facial-
recognition features offered by the state-of-the-art
cameras. This not only implies lower costs, but it also
means that by only performing algorithm improve-
ments, one can improve the quality of the solution
provided by the framework. Hence, hardware main-
tainance and support efforts are only limited to the
main functionality of the camera, i.e., capturing the
faces.
As can be observed in Section 3, the implemen-
tation of the entire framework requires merely a high
level of integration between the hardware and soft-
ware. This simplicity makes the framework even
more practical and hence desirable.
6 EVENT ORGANIZERS &
ATTENDEES
In a broader context, our framework can be extended
to be deployed by large event organizers, such as
Expos. Event organizers can attract more organiza-
tions around the globe by providing features using our
proposed framework. Additions can be easily made
to allow the framework to collect and analyze the
attendee-feedback data related to a particular industry
and share this data with the related companies. This
helps companies get insights not only about their own
products and presentations but also about other com-
petitors. Companies who wish to share such data and
be given such features could benefit from the experi-
ence of other companies and accordingly make their
future decisions.
Moreover, the framework could also be extended
to give insights to attendees. In huge events, it is com-
mon for attendees to visit several kiosks and keeping
track of the visits might not be easy. That is, an at-
tendee is very likely to forget his/her impression about
products or companies unless he/she makes an effort
to. The extended framework could provide each at-
tendee, upon his/her request, a summary of and an
overall impression about the kiosks visited, the pre-
sentations attended, and the interactions made during
the event. This makes the attendee’s visit more effi-
cient and fruitful.
7 CONCLUDING REMARKS &
FUTURE WORK
We have proposed in this paper a new AI-based
framework that leverages event marketing perfor-
mance by optimizing attendee-feedback generation
using a facial-recognition algorithm. This framework
ICAART 2022 - 14th International Conference on Agents and Artificial Intelligence
572
helps optimize event marketing strategies made by
companies after an event and during it.
A next step would be to implement this framework
in a real-world event and compare its performance to
that of a traditional framework. As described earlier,
the approach used in the framework is time-efficient
and can be easily implemented, making it a promising
event managing and marketing tool.
Like any other AI-based framework, it is impor-
tant to address the ethical concerns which accompany
the proposed framework (Liu et al., 2021; Zhu et al.,
2020; Oseni et al., 2021). While such events are gen-
erally public and it is common to have cameras in-
stalled for security purposes, it is still important to
get the approval of attendees when it comes to us-
ing their photos and analyzing them to get insights.
Hence, depending of the region where the event is
taking place, the proposed engagement measurement
framework will be accompanied with a number of
procedures, policies, and guidelines.
We have demonstrated in this study the role of AI
in leveraging event marketing performance, focusing
on attendee engagement as an indicator of the quality
of presentations/events. Our proposed platform ana-
lyzes the feedback of attendees by observing the at-
tendees’ facial expressions. The role of the platform
becomes even more vital in the case of presentations
in which there is no human presenter. In this case, the
system is expected to make an automatic intervention
as a reaction to the feedback gathered. To achieve
this, the system is fed with a number of presenting
formats such as image, video, or text. As a reaction
to the attendees’ feelings, the system would automat-
ically shift between these formats to produce the best
results.
REFERENCES
Campbell, C., Sands, S., Ferraro, C., Tsao, H.-Y. J.,
and Mavrommatis, A. (2020). From data to ac-
tion: How marketers can leverage ai. Business
Horizons, 63(2):227–243. ARTIFICIAL INTELLI-
GENCE AND MACHINE LEARNING.
De Bruyn, A., Viswanathan, V., Beh, Y. S., Brock, J. K.-
U., and von Wangenheim, F. (2020). Artificial intelli-
gence and marketing: Pitfalls and opportunities. Jour-
nal of Interactive Marketing, 51:91–105.
Ekman, P. (1993). Facial expression and emotion. American
psychologist, 48(4):384.
Findface. Face detection, verification and recognition tech-
nology. Available at https://findface.pro/en/.
Gupta, S. (2003). Event marketing: Issues and challenges.
IIMB Management Review, 15(2):87–96.
Hoyle, L. H. (2016). Event marketing. John Wiley & Sons,
Inc.
Huang, M.-H. and Rust, R. T. (2021). A strategic frame-
work for artificial intelligence in marketing. Journal
of the Academy of Marketing Science, 49(1):30–50.
Kietzmann, J., Paschen, J., and Treen, E. (2018). Artificial
intelligence in advertising: How marketers can lever-
age artificial intelligence along the consumer journey.
Journal of Advertising Research, 58:263–267.
Liu, B., Ding, M., Shaham, S., Rahayu, W., Farokhi, F., and
Lin, Z. (2021). When machine learning meets privacy:
A survey and outlook. ACM Comput. Surv., 54(2).
Neuhofer, B., Magnus, B., and Celuch, K. (2020). The
impact of artificial intelligence on event experiences:
a scenario technique approach. Electronic Markets,
pages 1–17.
Niedenthal, P. M., Halberstadt, J. B., Margolin, J., and
Innes-Ker,
˚
A. H. (2000). Emotional state and the de-
tection of change in facial expression of emotion. Eu-
ropean journal of social psychology, 30(2):211–222.
Oseni, A., Moustafa, N., Janicke, H., Liu, P., Tari, Z., and
Vasilakos, A. (2021). Security and privacy for artifi-
cial intelligence: Opportunities and challenges. arXiv
preprint arXiv:2102.04661.
Richhariya, B. and Gupta, D. (2019). Facial expression
recognition using iterative universum twin support
vector machine. Applied Soft Computing, 76:53–67.
Services, H. B. R. A. (2018). The event marketing evolu-
tion. Technical report.
Sterne, J. (2017). Artificial intelligence for marketing:
practical applications. John Wiley & Sons.
Whitehill, J., Serpell, Z., Lin, Y.-C., Foster, A., and Movel-
lan, J. (2014). The faces of engagement: Automatic
recognition of student engagement from facial expres-
sions.
Woodward Jr, J. D., Horn, C., Gatune, J., and Thomas,
A. (2003). Biometrics: A look at facial recognition.
Technical report, RAND CORP SANTA MONICA
CA.
Zhu, T., Ye, D., Wang, W., Zhou, W., and Yu, P. S. (2020).
More than privacy: Applying differential privacy in
key areas of artificial intelligence. arXiv preprint
arXiv:2008.01916.
Leveraging Event Marketing Performance using AI in Facial Recognition
573