/VGG16_RNN_LSTM_SignLanguageRecognition

The project of graduation essay.

Primary LanguageJupyter Notebook

Sign Language Recognition

Sections

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.