Uploaded by VAIBHAV SINGH

Group 8 mini project (1)

advertisement
Deep-emotion: Study of facial expression recognition
using attentional convolutional neural network
Guided By:
Proff. Arun Kumar Dubey
25th March 2022
Presented By:
Ishank
Vaibhav Singh
MINI PROJECT 2021
02611503119
01011503119
1
INTRODUCTION
The objective of this project is to develop a system which can analyze
the image and predict the expression of the person. This procedure is
workable and valid results can be produced.
The 7 emotions we will be classifying faces into are Anger, Disgust, Fear
, Happy , Sad , Surprise , Neutral
25th March 2022
MINI PROJECT 2021
2
Why emotion detection
- The motivation behind this topic specifically lies in the huge
investments large corporations do in feedbacks and surveys but fail to
get equitable response on their investment.
- Emotion detection through facial gestures is a technology that aims to
improve product and services performance by monitoring customer
behaviour to certain products or service staff by their evaluation.
25th March 2022
MINI PROJECT 2021
3
How to find facial expression
1- Vision based techniques
2 – Bio-Signals
- PPG,ECG,EEG
25th March 2022
MINI PROJECT 2021
4
Related work
1 - Earlier works on emotion recognition, rely on the traditional two-step machine
learning approach, where in the first step, some features are extracted from the images,
and in the second step, a classifier are used to detect the emotions.
2 - Limitations of previous approach
These approaches seemed to work fine on simpler datasets, but with the advent of
more challenging datasets (which have more intra-class variation), they started to
show their limitation.
25th March 2022
MINI PROJECT 2021
5
Review of dataset FER 2013
Problems in dataset
1- Imbalance problem
2- Intra-class variation of FER
3 - Contrast variation
4 – Eyeglasses
25th March 2022
MINI PROJECT 2021
6
Framework
Steps
1 – Data pre-processing
2 – Image augmentation
3- Feature extraction
4 – Training
5 – Validation
25th March 2022
MINI PROJECT 2021
7
Validation phase
Various open cv and keras functions have been used
We can pass the input from
- Any pre-existing image available in memory
- Live feed via system’s camera
The image frame is converted into grayscale and resized and reshaped with help of
numpy
A rectangle is drawn around the facial region and the output is formatted above the
rectangular box
25th March 2022
MINI PROJECT 2021
8
Accuracy of our model
1- Accuracy our model
Below are the few epochs of training process with batch size of 128
60%
25th March 2022
MINI PROJECT 2021
9
Download