Sign Language Detection

Project Description

Sign Language Detection is an AI enabled project aimed to address the crucial problem of communication gap between the people with mutism and the normal people. Though the project is very far from perfect, it's a tiny contribution from my side and a glimpse of what AI can do.

Key Concepts

• Image Recognition
• Semenatic Segmentation
• Convolutional Neural Network

Tools and libraries

• Language:- Python
• IDE:- PyCharm
• Libraries:- Keras, TensorFlow, OpenCV

Dataset

Training Dataset

• Total Classes:- 26 (Total Albhabets)
• Images per class:- 45500

Test Dataset

• Total Classes:- 26 (Total Albhabets)
• Images per class:- 6500

Algorithm

1. Gestures image caputed through Semantic Segmentation
2. Training a CNN neural network

Training Description

CNN (Convolutional Neural Network)

• Model:- Sequential, 3 Conv2D layer, 3 MaxPooling layer, 1 Flatten layer, 2 Dense and 1 Dropout layer
• Optimizer:- SGD(Stochastic Gradient Descent)
• Epochs:- 25

Screenshots

Prediction for 'B'	Prediction for 'C'

How to run the program?

Install required libraries

• Run pip install -r requirements.txt

Run the program

(Note:- Before running the program make sure to make your device stable)

• Step 1:- python run.py
• Step 2:- Let the program run first
• Step 3:- Do not change the position of the device once the program runs
• Step 4:- Give correspoding hand gestures for the letters (A- Z) and observe the result

Future Extension

• Training the network with words like Hello, Good and so on.
• Create a plugin for a messenger