CAPSTONE PROJECT

FACE EMOTION RECOGNITION

Batch ID : 10

Guide: Sangar Rajagopal

Batch Members
Binay Mohanty(7205788754)
Brajeswari Kabisatpathhy(6370921557)
Soumyanshu Sahoo(9090625821)
Sitanshu Malik(7609856178)
Subhrakanta Khandai(8260994314)
 MOTIVATION & OBJECTIVE

MOTIVATION:
 Enhance human-computer interaction.
 Use facial emotion analysis in areas like mental health, marketing, and
entertainment.
 Improve communication between humans and AI-powered systems by
enabling machines to recognize and respond to emotions.

OBJECTIVE:
 Build a model to classify emotions into six categories: Ahegao, Angry,
Happy, Neutral, Sad, and Surprise.
 Develop a deep learning-based system with high accuracy in recognizing
facial emotions under varying conditions (lighting, angle, etc.).

Foundation For Innovation and Technology Transfer, IIT,

Delhi 2
 INTRODUCTION
 What is Face Emotion Recognition?

 AI-powered system to analyze and classify human facial expressions.

 Relies on large labeled datasets of facial images to train and fine-tune
the recognition models.
 Utilizes deep learning and computer vision techniques to identify and
classify human facial expressions.
 Can be combined with audio or text-based sentiment analysis for
holistic emotion detection.

 Importance:
 Improves human-computer interaction.
 Applications in healthcare, security, and customer service.
 Creates personalized and adaptive experiences in applications like
gaming, virtual assistants, and e-learning platforms.

Foundation For Innovation and Technology Transfer, IIT,

Delhi
3
 LITERATURE REVIEW

Paul Ekman’s Research on Facial Expressions (1970s):

 Identified six universal emotions: Happiness, Sadness, Anger, Fear,
Surprise, and Disgust.
 Provided a theoretical foundation for facial expression analysis.
Traditional Approaches:
 Early methods used hand-crafted features such as Gabor filters, Local
Binary Patterns (LBP), and Support Vector Machines (SVM) for
classification.
 These approaches were limited by accuracy and struggled with variations
in lighting, pose, and occlusions.
Convolutional Neural Networks (CNNs):
 Revolutionized emotion recognition by automating feature extraction and
achieving state-of-the-art accuracy.
 Examples: AlexNet, VGGNet, and ResNet architectures applied to FER
datasets.

Foundation For Innovation and Technology Transfer, IIT,

4
Delhi
 DATASET USED

Dataset Details:
 Classes: Ahegao , Angry, Happy, Neutral, Sad, Surprise.
 Size: 15,453 images resized to 128x128 pixels.
 Datasets are collected from Kaggle.

Preprocessing:
 Image resizing and augmentation (rotation, shifting, flipping).
 One-hot encoding for labels.

Foundation For Innovation and Technology Transfer, IIT,

Delhi
5
 ARCHITECTURE DIAGRAM

Foundation For Innovation and Technology Transfer, IIT,

Delhi
6
 METHODOLOGY
1. Emotion Categories:
 Ahegao
 Angry
 Happy
 Neutral
 Sad
 Surprise

2. Data Collection:
 The dataset is organized into folders, where each folder corresponds to one emotion
category.
 Images are collected in various formats like .jpg and .png.

3. Preprocessing:
Image Resizing: All images are resized to 128x128 pixels for consistency in input dimensions.
Filtering: Only images with the required dimensions (128x128x3 for RGB) are
included.

Foundation For Innovation and Technology Transfer, IIT,

Delhi
7
 METHODOLOGY

One-Hot Encoding: Each emotion category is converted into a one-hot encoded label for
classification:
Ahegao → [1, 0, 0, 0, 0, 0]
Angry → [0, 1, 0, 0, 0, 0]
Happy → [0, 0, 1, 0, 0, 0]
Neutral → [0, 0, 0, 1, 0, 0]
Sad → [0, 0, 0, 0, 1, 0]
Surprise → [0, 0, 0, 0, 0, 1]

4. Data Augmentation:
 To improve generalization and handle limited data, the ImageDataGenerator is used for
augmentation:
 Rotation: Up to 10 degrees.
 Width/Height Shift: Up to 10% of the image.
 Horizontal Flip: Randomly flips the images.

5. Train- Test Field:

 The dataset is split into:80% for Training
 20% for Testing

Foundation For Innovation and Technology Transfer, IIT,

Delhi 8
 RESULT & DISCUSSIONS

Loss: 0.8504

Foundation For Innovation and Technology Transfer, IIT,

Delhi
9
 RESULT & DISCUSSIONS

Accuracy: 0.8412
Test Accuracy: 0.8411

Foundation For Innovation and Technology Transfer, IIT,

Delhi
10
 RESULT & DISCUSSIONS

Foundation For Innovation and Technology Transfer, IIT,

Delhi
11
 SYSTEM SPECIFICATION

HARDWARE SPECIFICATION:
RAM : 8 GB
HARD DISK : 256 GB SSD
PROCESSOR : Intel i5

SOFTWARE SPECIFICATION:
SOFTWARE TOOLS : Python 3.9, Google Colab
PROGRAMMING LANGUAGE: Python
USED LIBRARY1 : Seaborn
USED LIBRARY2 : Matplotlib
USED LIBRARY3 : TensorFlow/Keras
USED LIBRARY4 : Pillow, Numpy, Scikit-Learn
Foundation For Innovation and Technology Transfer, IIT,
Delhi
12
 CONCLUSION

 Effective Emotion Classification: The model accurately classifies facial

emotions (Ahegao, Angry, Happy, Neutral, Sad, Surprise) using deep
learning techniques, particularly ResNet-152 with transfer learning.
 High Accuracy: Strong performance on the test set with good accuracy
and low loss, validated by confusion matrix and classification report.
 Practical Applications: Potential use in healthcare, security, and customer
service for real-time emotion analysis.
 Future Enhancements: Opportunities to improve model performance
with more diverse datasets, better hyperparameter tuning, and real-time
optimization.
Foundation For Innovation and Technology Transfer, IIT,
Delhi
13
 SNAPSHOT

Foundation For Innovation and Technology Transfer, IIT,

Delhi
14
 SNAPSHOT

Foundation For Innovation and Technology Transfer, IIT,

Delhi
15
 REFERENCES

 https://www.deeplearningbook.org

 https://www.kaggle.com/datasets/msambare/fer2013

 https://papers.nips.cc/paper/2012/hash/6c1a3d57be94c8db9f0b68210d89fe21-

Abstract.html

 https://link.springer.com/chapter/10.1007/978-981-13-2513-0_20

 https://link.springer.com/article/10.1007/s12193-015-0186-2

 https://ieeexplore.ieee.org/document/7410701

 https://keras.io/

Foundation For Innovation and Technology Transfer, IIT,

Delhi
16