The document describes a project on sign language recognition using OpenCV by presenting the steps of the CNN algorithm used which includes data collection, image pre-processing, feature extraction, classification, and output of the recognized sign. It discusses achieving 90.44% accuracy for alphabets, 91.78% for numbers, and 88.90% for daily life symbols using a dataset of 8,800 images created for the project.
Seal of Good Local Governance (SGLG) 2024Final.pptx
epics.pptx
1. Sign Language Recognition using OpenCV
Presented by
Mr. Pavan Kumar Meka(208w1a12a8)
Mr. Yesu Raju Parusu(208w1a12c8)
Department of Information Technology
V R Siddhartha Engineering College
B.Tech in Information Technology
EPICS Project
Review 1
Under the guidance of
V. Radhesyam, Assistant Professor
Computer Vision
2. S.NO Type Symbol Meaning
Numbers
1 1
2 2
3 3
4 4
5 5
6 6
7 7
NUMBERS
3. S.NO Type Symbol Meaning
8 8
9 9
10 A
11 B
12 C
13 D
14 E
ALPHABETS
NUMBERS
4. S.NO Type Symbols Meaning
15 F
16 G
17 H
18 I
19 J
20 K
21 L
ALPHABETS
5. S.NO Type Symbols Meaning
22 M
23 N
24 O
25 P
26 Q
27 R
28 S
ALPHABETS
6. S.NO Type Symbol Meaning
29
ALPHABETS
T
30 U
31 V
32 W
33 X
34 Y
35 Z
7. S.NO Type Symbols Meaning
36
Daily life
symbols
Okay
37 Peace
38 Thumbs up
39 Thumbs down
40 Call me
41 Rock
10. CNN Algorithm Steps
• Step1 : Data Collection
It is a very crucial part of the research works in all the arenas as it is fundamental to
foster the development of any machine or deep learning model.
• Step2 : Image Pre-processing and segmentation
In this step, we step we are going to segment the image, separating the background from
foreground objects and we are going to further improve the quality of the image so that we can
analyze it in a better way.
Image Enhancement
• Step3 : Feature Extraction
Feature extraction refers to the process of transforming raw data into numerical features that
can be processed while preserving the information in the original data set. It yields better results
than applying machine learning directly to the raw data.
11. Algorithm Steps (Cont…)
• Step4 : Classification
4.1 Convolutional neural networks
CNNs are functional extraction models inspired by the human
brain’s visual cortex. CNNs compare images piece by piece
where a filter map slides over the local patches of the image.
Such pieces are called features, and they compare two images by
finding approximately the same features at approximately the
same locations.
CNNs have a better ability to see images and classify them than
other neural networks.
• Step5 : Output Sign
12. Accuracy
• Accuracy =
• 26 alphabets with recognition rate of 90.44% accuracy was
obtained.
• 9 numbers with recognition rate of 91.78% accuracy was
obtained.
• 9 daily life signals with recognition rate of 88.90% accuracy
was obtained.
Number of correct predictions
Total number of predictions
13. Data sets
• For this project we have created our own data sets
• Which contains 26 alphabetical images and each of 200 images
• And 9 numbers each of 200
• And 9 daily life symbols each of 200
• So totally we have created 8,800 images of dataset.