A C++ video analysis framework built as a pipeline of independent nodes, for running CV models and multimodal LLMs on video streams.
BIT-DataLab/Edit-Banana
Converts static images of diagrams, flowcharts and formulas into fully editable DrawIO (XML) files using SAM 3 segmentation and OCR.
- Stars
- 5,498
- Forks
- 0
- Open issues
- 0
- Last commit
- 3 days ago
Key features
- 01Segments diagram elements with a fine-tuned SAM 3 model.
- 02Exports reconstructed diagrams as editable DrawIO XML.
- 03Recognizes text with Tesseract or PaddleOCR and formulas as LaTeX via Pix2Text.
- 04Provides a command-line interface and a FastAPI web endpoint.
- 05Supports PNG, JPG, BMP, TIFF and WebP input images.
Use cases
- Turning scanned or exported flowcharts and architecture diagrams into editable drawings.
- Recovering editable formulas from figures in papers or slides.
- Batch converting legacy diagram images into DrawIO files for further editing.
Topics
FAQ
- What is Edit-Banana?
- Converts static images of diagrams, flowcharts and formulas into fully editable DrawIO (XML) files using SAM 3 segmentation and OCR.
- Is Edit-Banana free to use? What license does it have?
- Edit-Banana is open source under the AGPL-3.0 license. Check the license terms before using it in commercial products.
- Is Edit-Banana still maintained?
- Actively maintained. The last commit to Edit-Banana was on Oct 17, 2026.
- What can I use Edit-Banana for?
- Turning scanned or exported flowcharts and architecture diagrams into editable drawings. Recovering editable formulas from figures in papers or slides. Batch converting legacy diagram images into DrawIO files for further editing.
- Where can I get Edit-Banana?
- The source code and installation instructions are on GitHub: https://github.com/BIT-DataLab/Edit-Banana
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
Open-source Python OCR library that detects and recognizes text in PDFs and images using deep learning models, built on PyTorch.
Converts UI screenshots and design mockups into editable HTML/CSS code using a modular multi-agent pipeline.
A React and FastAPI web app for running DeepSeek-OCR on images and multi-page PDFs, with Markdown, HTML, DOCX, and JSON export.
Local-first video knowledge base that indexes your videos with transcription, frame analysis and embeddings, then lets you search scenes in natural language.
Python library and web UI that converts PDFs, DOCX, PPTX, XLSX, images and URLs into Markdown, JSON, CSV or HTML, with OCR and structured field extraction.