You are on page 1of 12

Facial Expressions Recognition

for Emotions Analysis


Group A1

Rishi Acharya
Roma Singh
Shikhar Shukla
Sujit Gaikwad
Domain :
Computer Vision

Application Domain :
Smart Cities

Research Themes :
Transportation
Motivations

● The way of interaction between human beings is more than just verbal communication. Humans rely
a lot on non-verbal methods of communication, specifically communication and understanding each
other through facial expressions.

● While verbal methods of communication may be primarily used, a lot can be derived from a person’s
non-verbal communication, particularly their expressions. Facial expressions are more explanatory
in situations where words fail, like a shock or a surprise. Moreover, lying through spoken words is
more difficult to detect compared to faking expressions.

● Statistically non-verbal forms of communication make up about 66% of the total communication.
This makes it very essential to study, use and analyze facial expressions.

● The use of analyzing facial expressions could be put to a variety of use cases. Consider a scenario
in which the system detects drowsiness of a driver in the car, and if detected, cause the alarm and
stop the car.
Research Papers and
References
● Research on Facial Expression Recognition Based on Neural Network 2020
ICCNEA
: Zhiheng Zhang Xi’an Shiyou University School of Electronic Engineering Control engineering Xi'an,
China E-mail: 763316846@qq.com
Ming Li Xi’an Shiyou University School of Electronic Engineering Control engineering Xi'an, China
E-mail: lm3007@126.com

● Analysis of Facial Emotion Recognition 2019 ICOEI


Balaji Balasubramanian Information Technology Vidyalankar Institute of Technology Mumbai, India
balajib26@gmail.com Rajeshwar Nadar Information Technology Vidyalankar Institute of Technology
Mumbai, India rajeshwar.nadar@vit.edu.in

● Emotion Recognition from Facial Expression using Deep Learning 2019 IJEAT
Nithya Roopa. S
Literature Surveys
Emotion Recognition from Facial Expression
using Deep Learning
Year Of Publication Author’s Name Conference/ Journal Gaps

August 2019 Nithya Roopa. S International Journal Because of transfer


of Engineering and learning the
Advanced accuracy generated
Technology (IJEAT) is 35.6%. It could’ve
generated less
number of
spectrograms for
training, which leads
to the less accuracy.
Literature Surveys
Year Of Publication Author’s Name Conference/ Journal Gaps

2019 ● Balaji International The paper doesn’t


Balasubraman Conference on describe the nature
ian Trends in of CNN in terms of
● Pranshu Electronics and emotion detections.
Diwan Informatics (ICOEI As its data points
● Anuradha 2019) are not defined in
Bhatia terms of emotion
detection. We need
large datasets for
classification and
difference in minute.
Literature Surveys
Year Of Publication Author’s Name Conference/ Journal Gaps

2020 Zhiheng Zhang International The Idea of Cross


Ming Li Conference on Connection in
Computer Network, AlexNet Structure
Electronic and Increases the
Automation training efficiency
(ICCNEA) but makes
recognition more
time-consuming, the
practically its not a
right fit for
applications that we
are trying to built.
Literature Gaps

After readings and analysis of 4 recent and 3 older papers, we have


found that the one time solutions for drivers FER is not addressed for
the sake of practical application.

Point-Gaps identified
1. Quickest analysis is required for such applications
2. A moderate accuracy is needed
3. The model should not be too complex by the nature of Pipe-line
Project Scope & Goals

We aim to built an Image Recognition Model FER which will have following
features.

● Lower latency for Emotion recognition.


● Eye Movement Tracking
● Easy Pipeline for Image filtering
● Specific for Application that we use.
● Can detect minute emotions like dizziness and fainting
● Requires less computational power
Hardware and Software
Requirements

Hardware Requirements

Camera : Pixy2
CPU : Arduino/ Raspberry
Installation setup : Device Holder

Software Requirements

Linux architecture in CPU


Trained and tested Model
Image Importing Pipe line
Technology Used

Use to CNN for Model training and data point Identification.


● Image Spectrum Rendering
● Alexnet
● Deep Learning Classifiers
Testing & Development
Strategies with Timeline.

Data set collection 20/09/2021

Pipeline building 1/10/2021

Algorithm Building -

Model testing -

You might also like