Are You Meme?

A fun project that matches your face and pose to popular memes using AI. Quickly find the meme that best represents your expression and share it with friends.

Demo

How it works

• Snap or upload a photo.
• Compute a compact visual fingerprint using multiple signals:
  • 🏃 Pose (40%): your body posture and position
  • ✋ Hand gestures (25%): thumbs up, peace sign, facepalm, etc.
  • 🖼️ Scene & vibe (20%): background/composition (faces mostly masked)
  • 🙂 Emotion (10%): facial expression
  • 📦 Objects (5%): props and items in frame
• Searches a FAISS index of meme embeddings and return the closest matches.
• Privacy: your photo is processed for matching and not stored.

Architecture

Embedding Generation (Offline)

Pre-process ~300 meme templates: extract CLIP, pose, hands, emotion, and object features. Concatenate into a 1223D vector per meme and build a FAISS index for fast similarity search.

Inference (Runtime)

When you upload a photo, we extract the same features, encode into a 1223D vector, and search the FAISS index using cosine similarity. The top matches are returned with scores.

Setup & Installation

Backend Setup

1. Navigate to the backend directory

   cd backend

2. Create a virtual environment (if it doesn’t already exist)

   python -m venv .venv

3. Activate the virtual environment

   source .venv/bin/activate

4. Install dependencies

   pip install -r requirements.txt

5. Run the embedding generation script

   python scripts/generate_embeddings.py

Start the Backend

1. Change to the backend directory and activate the virtual environment:

   cd backend
   source .venv/bin/activate

2. Run the FastAPI app with Uvicorn:

   uvicorn app:app --reload --host 127.0.0.1 --port 8000

Frontend Setup

1. Change to the frontend directory:

   cd frontend

2. Install dependencies if you haven't already:

   npm install

3. Start the development server:

   npm run dev

The frontend will be accessible at http://localhost:5173 by default.