Meme Explainer — Kaggle Gen AI Project

Course Participation

• I participated in the 5-day Gen AI Intensive Course with Google in April 2025 and built this project as part of my learning.
• I successfully obtained the certificate (badge): https://www.kaggle.com/certification/badges/monapouresmaeil/96

Project Overview

• Goal: Explain memes like a human by extracting their tone, humor type, and core message.
• Input: Meme image (+ caption from dataset).
• Output: Structured JSON (tone, humor, message), saved to CSV for analysis.

How It Works

• Load labels.csv from the Kaggle meme dataset and resolve image paths.
• Open each image and send it with a clear prompt to a vision-capable generative AI model.
• Parse the model’s response (prefer JSON) and collect results.
• Save explanations to meme_explanations.csv.

Tech Stack

• Data: pandas, numpy
• Images & Visualization: Pillow (PIL), matplotlib
• AI: Google Generative AI Python SDK
• Secrets: kaggle_secrets (to securely load GOOGLE_API_KEY)

Files

• final-competition-mona (1).ipynb — main notebook pipeline.
• explain.txt — complete conceptual + technical guide.

Run Notes

• Set GOOGLE_API_KEY (Kaggle secret or environment variable) before creating the model.
• Verify dataset paths (Kaggle: /kaggle/input/... or local: data/...).
• Run the notebook to generate meme_explanations.csv.