Hugging Face's logo

Spaces:
dsgn2stdy3rsrch-ai
/
test
Sleeping

App Files Community
test / README.md
munyakarg
run w/ docker and flack
4adbbc7
preview code
|
raw
history blame contribute delete
2.47 kB
metadata 
title: Test
emoji: 😻
colorFrom: purple
colorTo: indigo
license: apache-2.0
sdk: docker
app_port: 7860
python_version: 3.11.9

AI Filter & Demographic Detection Study

This application is part of a research study examining how image filters affect AI model predictions of demographic characteristics.

Features

  • Real-time image processing with filters
  • AI-powered demographic detection
  • Anonymous data collection for research
  • Optional results delivery via email/SMS

Beauty-Filter-Smart-Mirror-Audit

This repository combines an interactive Smart Mirror web application with a Python analysis pipeline to measure how Instagram and IG-like beauty filters alter facial features, skin tone, and downstream predictions of race, age, and gender.

Running the Beauty Filter Smart Mirror Application

  1. Navigate to the project root directory cd /Beauty-Filter-Smart-Mirror-Audit

  2. Create a virtual environment (if not already created) python3 -m venv venv

  3. Activate the virtual environment source venv/bin/activate

  4. Install all required dependencies pip install -r requirements_all.txt

  5. Run the application from the project root /Beauty-Filter-Smart-Mirror-Audit

  6. Start the Flask application from the root directory: python app/app.py

read more about how to setup the environment in: SETUP.txt

Important Note

The application must be run from the project root directory (Beauty-Filter-Smart-Mirror-Audit). The code relies on relative file paths to load the TensorFlow Lite models stored in the models/ directory. Running the app from another directory may result in model-loading errors.

Folder Structure and .keep Files

This repository uses empty .keep files to preserve directory structure in Git.

Git does not track empty folders by default. Some directories in this project (e.g., data/ and models/) are intentionally kept empty in the repository because they are populated locally with large files (datasets or model weights) that should not be committed.

The .keep files allow these folders to exist consistently across branches and merges without tracking their contents.

Downloading Data (Dropbox)

Large files are not stored in the repo. Use the download_data script:

pip install requests

Download multiple ZIPs at once:

python scripts/download_data.py \
  --dropbox_urls "<LINK_1>" "<LINK_2>" \
  --output_dir data/ \
  --zip_name dataset