Skip to content

EricAzayev/CornellAIHackathon

Repository files navigation

TruthLens

Product Overview

Our solution is a browser extension / platform plugin that detects and flags harmful or low-quality content on short-form video platforms, with initial support for Instagram Reels and/or YouTube Shorts.

The tool can be toggled on or off by the user. It uses feature extraction techniques to analyze signals from the video, audio, metadata, and engagement patterns, to predict whether AI-generated content is appropriate for younger audiences. With AI-generated videos becoming increasingly prevalent on social media, children are exposed to a growing volume of synthetic content that can be misleading, age-inappropriate, or simply unvetted. Our tool helps parents and guardians regulate that exposure by automatically identifying and flagging AI-generated Reels or Shorts reaching young viewers.

Members:

  • Eric Azayev
  • Samira Maria
  • Neha Sudarshan
  • Preet Patel

System Architecture

System Architecture

Hackathon Setup Guide

1. Backend

cd backend
cp .env.example .env       # edit if needed
docker-compose up -d       # starts API + worker + Redis + Postgres

API will be live at http://localhost:8000 Docs at http://localhost:8000/docs

Test it:

curl -X POST http://localhost:8000/api/v1/analyze/sync \
  -H "Content-Type: application/json" \
  -d '{"media_url": "https://upload.wikimedia.org/wikipedia/commons/thumb/4/47/PNG_transparency_demonstration_1.png/280px-PNG_transparency_demonstration_1.png", "media_type": "image"}'

2. Chrome Extension

  1. Open Chrome → chrome://extensions
  2. Enable Developer Mode (top right)
  3. Click Load Unpacked
  4. Select the extension/ folder
  5. Browse to any image-heavy page — badges will appear

Architecture

Browser Extension (content.js)
    ↓ POST /api/v1/analyze/sync
FastAPI (main.py)
    ↓ runs inference
HuggingFace Model (umm-maybe/AI-image-detector)
    ↓ returns score
Extension injects badge with verdict

Swapping the model

Edit backend/app/services/detector.py, change MODEL_ID:

# Good alternatives to benchmark:
# "Organika/sdxl-detector"          — better on Stable Diffusion outputs
# "haywoodsloan/ai-image-detector"  — ensemble approach
# "prithivMLmods/Deep-Fake-Detector-Model" — face-focused

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors