LinguaAR is a collaborative Filipino Sign Language learning and accessibility prototype. Its source combines a Flutter learning interface with a Python gesture-processing service and supporting speech and augmented-reality packages.
My role and the evidence scope#
I present LinguaAR as a project contributor, with this study focused on the inspected application and gesture pipeline. The repository contains work by multiple contributors; this page does not claim sole authorship of the application or trained model.
The learning concept includes lessons, quizzes, flashcards, and favorites. The inspected Flask route, Flutter repository, and package definitions establish implementation evidence rather than completed user-study results.
Follow a hand image through the service#
The gesture route detects a hand, crops its bounding region with padding, and resizes it into a 300-by-300 white canvas while preserving aspect ratio. A Keras classifier selects from a fixed set of 24 static labels.
That path classifies an image. It does not establish continuous FSL translation, sentence interpretation, or measured recognition accuracy. The inspected label set omits J and Z.
Make integration limits explicit#
The inspected Flutter hand repository requests a different endpoint and multipart field from the examined detection route. The source therefore shows an integration contract that needs alignment or verification before describing that pair as a working end-to-end flow.
Result and next evidence#
The public source makes the prototype's learning and classification architecture reviewable. A complete evaluation would identify the deployed route, tested client contract, model data, label coverage, and actual test conditions. This study reports no invented accuracy or learner outcomes.
Read the companion article on the hand-image classification pipeline.