README.md

Sign Language Recognition

Sections

• Description
• Dataset
• Architecture
• Methods
• Result
• Getting Started

Description

The project of graduation essay.

In this work, we have applied Deep Learning for sign language recognition. VGG16, RNN and LSTM were used.

Member

• Vu Truong Giang
• Tat Tran Phong

Dataset

We use LSA64: A Dataset for Argentinian Sign Language
It is available on Link.

Architecture

Null

Methods

VGG16 + LSTM method:

VGG16_LSTM_Train : file for training by LSTM model

CM_VGG16_LSTM_Test : file for testing LSTM model on test set

VGG16 + RNN method:

VGG16_RNN_Train : file for training by RNN model

CM_VGG16_RNN_Test : file for testing RNN model on test set

Result

Index	Name	Accuracy
1	VGG16 + RNN	82.81%
2	VGG16 + LSTM	95.62%

Getting Started

Library: NumPy, os, Matplotlib, Tensorflow, Keras, Sklearn, opencv, Pandas, Seaborn

LSTM model

• Run file VGG16_LSTM_Train.ipynb to train data and create weights.
• Run file CM_VGG16_LSTM_Test.ipynb for testing model and showing result.

RNN model

• Run file VGG16_RNN_Train.ipynb to train data and create weights.
• Run file CM_VGG16_RNN_Test.ipynb for testing model and showing result.