Scribd
Upload
63 views11 pages

NSU - CSE465 - Project Analysis

The document describes a student project on facial emotion detection using convolutional neural networks (CNN) and attention mechanisms. It details the datasets and models used, including training a CNN and attention neural network on the CK+ dataset of facial images labeled with eight emotions. Hardware usage and the number of parameters and computations for each model are provided. Test results including loss, accuracy, and performance metrics are reported for both models.

Uploaded by

AkilProttoy
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
63 views11 pages

NSU - CSE465 - Project Analysis

The document describes a student project on facial emotion detection using convolutional neural networks (CNN) and attention mechanisms. It details the datasets and models used, including training a CNN and attention neural network on the CK+ dataset of facial images labeled with eight emotions. Hardware usage and the number of parameters and computations for each model are provided. Test results including loss, accuracy, and performance metrics are reported for both models.

Uploaded by

AkilProttoy
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 11

Assignment 01

Name: Akil Hossain Prottoy


ID:1931795642
Section: 03
Topic: Facial Emotion Detection
Using CNN and Attention Mechanism
Submitted To: Ashrafuzzaman Khan
(AZK)
Submission Deadline:12/11/2023
1) The project was about detecting facial emotion using Convolutional Neural Network and
Attention Mechanism in CNN. Models were created for both the networks. And also trained and
tested successfully.

2) CK+ dataset was used. It stands for Extended Cohn-Kanade dataset and it taken from Kaggle’s
dataset repository. The dataset contains 981 images and is divided into 8 classes, each class
representing a particular emotion. The emotions are:
0 : Anger
1 : Disgust
2 : Fear
3 : Happiness
4 : Sadness
5 : Surprise
6 : Neutral
7: Contempt

3) Hardware Usage:

For CNN:

Number Of CPU usage CPU usage Memory Memory


Epochs before after training usage before usage after
training training training

1 2.7% 78.1% 45.3% 45.1%


2 0.0% 88.7% 45.1% 45.0%
3 0.0% 89.1% 45.0% 45.0%
4 0.0% 88.0% 45.0% 45.1%
5 0.0% 89.7% 45.1% 46.1%
6 0.0% 89.4% 46.1% 46.7%
7 0.0% 88.0% 46.7% 46.1%
8 0.0% 87.8% 46.1% 46.0%
9 0.0% 88.8% 46.0% 46.4%
10 0.0% 88.3% 46.4% 46.5%
For Attention Neural Network:

Number Of CPU usage CPU usage Memory Memory


Epochs before after training usage before usage after
training training training

1 5.7% 61.9% 5.0% 6.7%


2 0.0% 64.7% 6.7% 7.3%
3 0.0% 61.4% 7.3% 8.0%
4 0.0% 61.2% 8.0% 8.4%
5 0.0% 62.0% 8.4% 8.8%
6 0.0% 62.8% 8.8% 9.3%
7 0.0% 61.7% 9.3% 9.6%
8 0.0% 64.1% 9.6% 10.0%
9 0.0% 61.1% 10.0% 10.3%
10 0.0% 60.9% 10.3% 10.6%

4) The number of parameters and number of computations for the forward pass of both models
are given below:

For CNN:
Total Number of Parameters: 217983
Number of Computations for a Forward Pass: 32812545024
Model Summary:
For Attention Neural Network:
Total Number of Parameters: 1553033
Number of Computations for a Forward Pass: 1846717440

Model Summary:
5) Metrics are given below:

For CNN:
Test Loss: 1.3095649480819702
Test accuracy: 0.4771573543548584
(1, 48, 48, 3)
Training and Validation curves:
Confusion matrix:

Performance Metrics:
Accuracy: 0.48
Precision: 0.41
Recall: 0.48
For Attention Neural Network:

Test Loss: 0.3650


Test accuracy: 0.8901

Training and Validation Curves:


Confusion Matrix:

Performance metrics:
Accuracy: 0.89
Precision: 0.93
Recall: 0.89

You might also like