Meme Classification Pipeline

Overview

This repository hosts a comprehensive Meme Classification Pipeline designed to determine whether a given meme contains harmful or hateful content. Leveraging the power of gpt 3.5 turbo and a suite of tools, the pipeline provides an effective mechanism to assess and categorize memes based on their textual and visual content.

Features

Text Extractor:

Extracts textual content from meme images.

Image Description Tool:

Generates contextual descriptions of images to capture the essence and nuances.

How It Works

Text Extraction: The pipeline begins by extracting text from meme images using advanced text extraction techniques. Image Description: An Image Description Tool processes the images to generate meaningful contextual descriptions. Contextual Analysis: The extracted text and image descriptions are fed into the LLM for a thorough analysis. Classification: Based on the analysis, the pipeline classifies memes into categories such as harmful (hateful content) or benign.