Home

Awesome

ABHINAW-MATRIX

We propose a novel evaluation matrix for assessing text and typography in AI-generated images, addressing gaps in current benchmarking methods like CLIP SCORE and T2I-CompBench++. Our matrix uses automated tools for reproducible benchmarks, setting a new standard in text accuracy evaluation for AI image synthesis platforms. Overview Welcome to the Abhinaw Matrix project! This repository contains tools and instructions to evaluate the accuracy of text and typography in AI-generated images using our novel evaluation matrix.

Table of Contents Getting Started Prerequisites Step-by-Step Guide

  1. Generate Images
  2. Use Abhinaw Matrix
  3. Extract Data
  4. Save Data
  5. Gauge Exact Score Contributing License Getting Started To get started with the Abhinaw Matrix, follow the steps below to set up your environment and run the evaluation on your AI-generated images.

Prerequisites Ensure you have the following installed:

Python 3.x

Step-by-Step Guide

  1. Generate Images Generate images using your preferred AI text-to-image generation platform (e.g., MidJourney, DALL-E, Stable Diffusion).

  2. Use Abhinaw Matrix Use the Abhinaw Matrix tool available in GPT STORE to evaluate the generated images. Link- https://chatgpt.com/g/g-TETU7gI79-abhinaw-matrix

  3. Extract Data Extract the evaluation data in tabular format from the Abhinaw Matrix tool.

  4. Save Data Save the extracted data in CSV format.

  5. Gauge Exact Score Use the provided Python code in this repository to gauge the exact score for your data.

Contributing We welcome contributions! Please see the CONTRIBUTING.md for more details on how to contribute.

License This project is licensed under the Apache License - see the LICENSE file for details.