ASL Sign Language Detector (local)

Overview

This project runs an ASL sign detector locally. It streams webcam frames to a Flask app (/video_feed) and recognizes letters and its given accuracy. Future imrprovement ideas include using the OpenAPI to construct complete words and also adding support for numbers.

Files

• app.py — Flask app that streams /video_feed and exposes /current for the corrected word.
• keras_Model.h5 — your trained Keras model (place the real file here).
• labels.txt — newline-separated class labels (A-Z or custom labels).
• templates/index.html — frontend page that displays the video.

Setup

1. Create and activate a Python virtual environment (recommended).

Windows PowerShell example:

python -m venv .venv; .\.venv\Scripts\Activate.ps1
pip install -r requirements.txt

2. Place your keras_Model.h5 and labels.txt in the project root.

3. (Optional) Configure OpenAI for spelling correction:

setx OPENAI_API_KEY "your_api_key_here"
# Then restart your shell/IDE so the env var is visible to processes

Run

Start the Flask app:

python app.py

Open http://localhost:5000/ in your browser. The video stream is at /video_feed.

Notes

• The app will not send any other user text to GPT; only the ASL tentative word is sent for correction.
• If you run into model import errors, ensure tensorflow is installed and your Python version is compatible.
• For production or remote camera use, modify the cv2.VideoCapture() source and secure the OpenAI key handling.