Document OCR using Nanonets-OCR-s

Name: my-ocr-output
Creator: mak

This dataset contains markdown-formatted OCR results from images in /content/my_dataset using Nanonets-OCR-s.

Processing Details

Source Dataset: /content/my_dataset
Model: nanonets/Nanonets-OCR-s
Number of Samples: 1
Processing Time: 4.6 minutes
Processing Date: 2025-08-11 09:33 UTC

Configuration

Image Column: image
Output Column: markdown
Dataset Split: train
Batch Size: 1
Max Model Length: 8,192 tokens
Max Output Tokens: 4,096
GPU Memory Utilization: 80.0%

Model Information

Nanonets-OCR-s is a state-of-the-art document OCR model that excels at:

📐 LaTeX equations - Mathematical formulas preserved in LaTeX format
📊 Tables - Extracted and formatted as HTML
📝 Document structure - Headers, lists, and formatting maintained
🖼️ Images - Captions and descriptions included in <img> tags
☑️ Forms - Checkboxes rendered as ☐/☑
🔖 Watermarks - Wrapped in <watermark> tags
📄 Page numbers - Wrapped in <page_number> tags

Dataset Structure

The dataset contains all original columns plus:

markdown: The extracted text in markdown format with preserved structure
inference_info: JSON list tracking all OCR models applied to this dataset

Usage

from datasets import load_dataset
import json

# Load the dataset
dataset = load_dataset("{output_dataset_id}", split="train")

# Access the markdown text
for example in dataset:
    print(example["markdown"])
    break

# View all OCR models applied to this dataset
inference_info = json.loads(dataset[0]["inference_info"])
for info in inference_info:
    print(f"Column: {info['column_name']} - Model: {info['model_id']}")

Reproduction

This dataset was generated using the uv-scripts/ocr Nanonets OCR script:

uv run https://huggingface.co/datasets/uv-scripts/ocr/raw/main/nanonets-ocr.py \
    /content/my_dataset \
    <output-dataset> \
    --image-column image \
    --batch-size 1 \
    --max-model-len 8192 \
    --max-tokens 4096 \
    --gpu-memory-utilization 0.8

Performance

Processing Speed: ~0.0 images/second
GPU Configuration: vLLM with 80% GPU memory utilization

Generated with 🤖 UV Scripts