Skip to content

Repository dedicated to the research on Costa Rican sign language (LESCO) recognition. This could also be used for other sign language recognition

Notifications You must be signed in to change notification settings

drzamoramora/LESCO-recognizer

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

16 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

LESCO-recognizer

Summary:

This project is an evolution of Juan Zamora's Doctoral dissertation: "Video-Based Costa Rican Sign Language Recognition for Emergency Situations" from Aspen University. The main idea is to create a recognition framework that is able to "understand" LESCO from videos and be able to translate each video into a set of keywords in spanish (or any language) for full video-2-text translation.

Dataset:

The dataset has been uploaded to Zenodo

The dataset is composed of 39 signs. There are three videos for each sign on each folder. Videos have been cropped and are on average 1 second long. This dataset contains a total of 117 videos and 20 additional videos of LESCO sentences.

Methodology:

Design Science (DS) has been used as the main research methodology. DS provides a set of practices to translate an idea into a product leveraged into a set of iterations where every artifact is tested and evaluated for further iterations. Each iteration could be seen as an "Agile" iteration where ne wideas and hypothesis are tested.

Itertations

  • Iteration 1: translate LESCO videos into text by using similary measures
  • Iteration 2: translate LESCO videos into text by using Deep Learning
  • Iteration 3: evolution of Iteration 1 with other dimensional reduction algoritms.
  • Iteration 4: cherry-picked frames for each video were selected to reduce the amount of frames for training (from each video): the hypothesis is that key frames that show relevant hand shapes are sufficient for sign recognition.
  • Contie2022: Iteration 3 code-base for the paper "Costa Rican Sign Language Recognition Using MediaPipe"

Acknowledgements

  • This project has been actively supported with LESCO translators, and research guidance over inclusive technologies by IncluTec from the Intituto Tecnologico de Costa Rica and the school of Computer Science of the Universidad Latina de Costa Rica.

About

Repository dedicated to the research on Costa Rican sign language (LESCO) recognition. This could also be used for other sign language recognition

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published