toshisakao/handwritten_to_md

Python

8 commits

updated Feb 16, 2026

See the code

README

handwritten_to_md

Converts handwritten PDF notes into clean Markdown, with automatic figure/diagram extraction.

Python 3.12

Dependencies

System

poppler - required by pdf2image for PDF rendering
- Ubuntu/Debian: sudo apt install poppler-utils
- Arch: sudo pacman -S poppler
- macOS: brew install poppler
Ollama - local LLM runtime for OCR (text transcription)
- https://ollama.com
- Pull the OCR model: ollama pull qwen3-vl
NVIDIA GPU with CUDA support (tested on RTX 3070 Ti 8GB)

Python

Install all Python dependencies:

pip install -r requirements.txt

Key packages:

Package	Purpose
`torch`, `torchvision`	PyTorch backend (CUDA)
`transformers`	Florence-2 model for figure detection & verification
`timm`, `einops`	Required by Florence-2 architecture
`ultralytics`	YOLOv12 DocLayNet for layout detection
`huggingface_hub`	Model downloading (YOLO weights, Florence-2)
`ollama`	Qwen3-VL for handwritten text OCR
`opencv-python`	Image preprocessing (CLAHE, thresholding, denoising)
`pdf2image`	PDF page to image conversion
`pillow`	Image handling
`tqdm`	Progress bars
`numpy`	Array operations

Models (downloaded automatically on first run)

Model	Source	Size	Purpose
YOLOv12m-DocLayNet	`hantian/yolo-doclaynet`	~50MB	Fast layout detection (pass 1)
Florence-2-base	`microsoft/Florence-2-base`	~0.5GB	Figure detection, verification & crop refinement (pass 2)
Qwen3-VL	Ollama (`qwen3-vl`)	~5GB	Handwritten text OCR

Usage

Place PDF files in ./test/
Run:

python main.py

Output Markdown and extracted figures are saved to ./test_out/

Similar repos

fcomendoza-site/page2md

Page-level document parsing to Markdown with OvisOCR2 — runs locally on Apple Silicon (MLX), ~1.9GB…

cast42/pdf2md_lighton

Convert PDF to markdown with lightonai/LightOnOCR-2-1B

leoneversberg/pdf2md_llm

Use a local LLM to convert PDF to Markdown

cqptomii/TexFromPDF

MarkPDFdown/markpdfdown

A high-quality PDF to Markdown tool based on large language model visual recognition.…

2,297

aiminnovations/_aim_olmocr

Toolkit for linearizing PDFs for LLM datasets/training

allenai/olmocr

Toolkit for linearizing PDFs for LLM datasets/training

19,751

datalab-to/marker

Convert PDF to markdown + JSON quickly with high accuracy

40,548

toshisakao/handwritten_to_md

Python

8 commits

updated Feb 16, 2026

See the code

README

handwritten_to_md

Converts handwritten PDF notes into clean Markdown, with automatic figure/diagram extraction.

Python 3.12

Dependencies

System

poppler - required by pdf2image for PDF rendering
- Ubuntu/Debian: sudo apt install poppler-utils
- Arch: sudo pacman -S poppler
- macOS: brew install poppler
Ollama - local LLM runtime for OCR (text transcription)
- https://ollama.com
- Pull the OCR model: ollama pull qwen3-vl
NVIDIA GPU with CUDA support (tested on RTX 3070 Ti 8GB)

Python

Install all Python dependencies:

pip install -r requirements.txt

Key packages:

Package	Purpose
`torch`, `torchvision`	PyTorch backend (CUDA)
`transformers`	Florence-2 model for figure detection & verification
`timm`, `einops`	Required by Florence-2 architecture
`ultralytics`	YOLOv12 DocLayNet for layout detection
`huggingface_hub`	Model downloading (YOLO weights, Florence-2)
`ollama`	Qwen3-VL for handwritten text OCR
`opencv-python`	Image preprocessing (CLAHE, thresholding, denoising)
`pdf2image`	PDF page to image conversion
`pillow`	Image handling
`tqdm`	Progress bars
`numpy`	Array operations

Models (downloaded automatically on first run)

Model	Source	Size	Purpose
YOLOv12m-DocLayNet	`hantian/yolo-doclaynet`	~50MB	Fast layout detection (pass 1)
Florence-2-base	`microsoft/Florence-2-base`	~0.5GB	Figure detection, verification & crop refinement (pass 2)
Qwen3-VL	Ollama (`qwen3-vl`)	~5GB	Handwritten text OCR

Usage

Place PDF files in ./test/
Run:

python main.py

Output Markdown and extracted figures are saved to ./test_out/

Similar repos

fcomendoza-site/page2md

Page-level document parsing to Markdown with OvisOCR2 — runs locally on Apple Silicon (MLX), ~1.9GB…

cast42/pdf2md_lighton

Convert PDF to markdown with lightonai/LightOnOCR-2-1B

leoneversberg/pdf2md_llm

Use a local LLM to convert PDF to Markdown

cqptomii/TexFromPDF

MarkPDFdown/markpdfdown

A high-quality PDF to Markdown tool based on large language model visual recognition.…

2,297

aiminnovations/_aim_olmocr

Toolkit for linearizing PDFs for LLM datasets/training

allenai/olmocr

Toolkit for linearizing PDFs for LLM datasets/training

19,751

datalab-to/marker

Convert PDF to markdown + JSON quickly with high accuracy

40,548