From 1517a5910aa3be6ef39c4f30546a435368c33f4f Mon Sep 17 00:00:00 2001 From: Odunayo Ogundepo Date: Sun, 25 Jan 2026 14:37:15 +0000 Subject: [PATCH] Update README.md --- README.md | 35 +++++++++++++++++++++++++++++++++++ 1 file changed, 35 insertions(+) diff --git a/README.md b/README.md index d89c121..94b28a2 100644 --- a/README.md +++ b/README.md @@ -27,6 +27,26 @@ tags: # KarantaOCR: Efficient Document Processing for African Languages +## Table of Contents + +- [Model Description](#model-description) +- [Training Data](#training-data) + - [Stage 1: General OCR Training](#stage-1-general-ocr-training) + - [Stage 2: African Language Fine-Tuning](#stage-2-african-language-fine-tuning) + - [Training Plots](#training-plots) +- [Capabilities](#capabilities) +- [Evaluation](#evaluation) + - [Results -- KarantaOCR-Bench](#results----karantaocr-bench) + - [Results -- OlmoOCR-Bench](#results----olmocr-bench) +- [How to Use](#how-to-use) + - [Load the Model and Processor](#load-the-model-and-processor) + - [Prepare a PDF Page for Inference](#prepare-a-pdf-page-for-inference) + - [Run OCR Inference](#run-ocr-inference) + - [End-to-End Example](#end-to-end-example) +- [Citation Information](#citation-information) + +--- + ## Model Description #### [Paper](....) @@ -59,6 +79,21 @@ KarantaOCR was trained using a **two-stage curriculum fine-tuning strategy**. This stage emphasizes accurate transcription of **diacritics, special characters, and region-specific typography**. +### Training Plots + +
+Train Loss +
+ +
+Eval Loss +
+ +
+Learning Rate +
+ + --- ## Capabilities