👋 Welcome to the VLM OCR Competition!

This competition aims to improve **open-source Arabic OCR models**. It's part of the NAMAA Community mission to strengthen the Arabic presence in the ML space. This competition is designed to **push the boundaries** of OCR performance on diverse Arabic documents.

QARI OCR Competition Rules

Welcome to the QARI OCR Competition organized by the NAMAA Community and sponsored by KANDCA! The competition runs from September 15 to October 15. Join the Discord server for support and discussion. Full rules and submission portal: Hugging Face Space.

📜 Submission Rules

  • Each team can submit one model evaluation per week.
  • Provide with your submission:
    1. Team name (must stay consistent across submissions)
    2. Model name & Hugging Face Model ID
    3. A valid Hugging Face token with access
    4. The inference code and any dependency installation instructions
    5. The OCR output must be a single-page structured HTML using the following tags:
      • <header>, <footer>, <main>, <section id="1">, <section id="2">
      • <p>, <h1>-<h5>, <b>, <i>, <u>
      • <img>, <table>, <hr>, <ul>, <ol>
    6. Submitting only unstructured output will result in a 5-point deduction from your final score.
  • The submitted code is the responsibility of the submitting team.
  • Ideally, provide a working Google Colab link with all details and dependencies.

📆 Evaluation Schedule

  • Submissions received by Sunday at midnight will be evaluated on Monday.
  • The leaderboard will be updated by Wednesday or Thursday of the same week.