Skip to content

Latest commit

 

History

History
26 lines (21 loc) · 1.61 KB

README.md

File metadata and controls

26 lines (21 loc) · 1.61 KB
Raw

Conversion of Pakistan Sign Language into text and speech using Machine Learning

The University of Lahore (Department of CS & IT) (Fall 2015 – Spring 2019) - Final Year Project

Supervised by:

Dr. Mudasser Naseer

Group Members
  • Muhammad Junaid Ejaz
  • Muhammad Saad Qadri

The project is divided into 3 main modules

  • RealTime sign detection module - convert Pakistan Sign Language (PSL) alphabet and words into text and speech in realtime.
  • Capture Dataset module - Automated system for capturing and adding new data to the dataset
  • PSl learning module - Learn PSL interactively

Dependencies

The system uses OpenPose v1.5.0 library for extacting skelatal features.
The code was developed with python 3.7 and has been tested with the libraries/versions in requirements.txt file.

Dataset Used

We have made our own Pakistan Sign Language (PSL) dataset containing multiple samples of 37 urdu aplhabets and 12 urdu words. The dataset is made publically available at: https://www.kaggle.com/saadbutt321/pakistan-sign-language-dataset

External resources

OpenPose GitHub repo: https://github.com/CMU-Perceptual-Computing-Lab/openpose Origin of OpenPose: https://github.com/ZheC/Realtime_Multi-Person_Pose_Estimation Paper describing the method: https://arxiv.org/abs/1611.08050 Keras implementation of the Realtime Multi-Person Pose Estimation (my major inspiration): https://github.com/michalfaber/keras_Realtime_Multi-Person_Pose_Estimation.