Signo-Lingo

Signo Lingo is a video classification project that classifies sign languages into English semantic meaning. The model is trained on the Turkish Sign Language dataset (AUTSL).

1. Table of contents

1. Table of contents
2. Directory structure
3. Problem Statement

2. Directory structure

.
├── src/                                   # logs gathered from 
│   ├── dataset/                           # stores train,val,test dataset
│   ├── dev_model/                         # stores model
│   ├── test_model/                        # store notebook for testing
│   ├── train_final/                       # final notebook
│   ├── Exploratory Data Analysis.ipynb    # notebook to run data analysis
│   └── log_file_plots.ipynb               # code for plotting graphs based on log files
├── models/                                # saved models
├── web_gui/                               # code for web GUI
├── Big_Project_Instructions.pdf           # instructions for project
└── README.md

3. Problem Statement

Sign language is a way of expressing oneself primarily through a series of hand gestures. However, it is dominantly used by the deaf and mute community and may prove to be a challenge for others who do not use the language to understand. Moreover, sign language is not universal - each community develops its own sign language with unique intricacies. As such, this project aims to develop a sign language video classification model that will help others to interpret the semantic meanings behind the language.

Inputs: Videos of gestures representing words in the Turkish Sign Language.

Outputs: Classes that are mapped to a specific meaning, in English.

Deliverables: Source Code (models, weights, logs, notebooks) and GUI