[ICDAR 2023] SelfDocSeg: A self-supervised vision-based approach towards Document Segmentation (Oral)
-
Updated
Oct 6, 2023 - Python
[ICDAR 2023] SelfDocSeg: A self-supervised vision-based approach towards Document Segmentation (Oral)
A script that utilises Facebook's SAM-2 model to add segmentation mask, bounding box, and rotation angle annotations to the MIDV500 and MIDV2019 datasets.
Mapping a set of zones generated by a segmentation algorithm to the regions generated by OCR engine
Document Segmentation project in Python.
High-precision document boundary segmentation & perspective dewarping using YOLO11-segmentation. Features custom data engineering with zero data leakage (99.4% mAP50).
彝文古籍整页扫描自动切字:桌面复核程序(exe)+ 规范字库校验 + 字库管理程序 + 批量 CLI。OpenCV 纯 CPU,6 页手工框 mean F1 0.8768。
High-fidelity OCR + pre-RAG pipeline processor featuring: 1.) Tesseract OCR 2.) Built-in cross-line dehyphenation + real word verification 3.) Support for TIFF series, & JPEG2000 (jpx) for hi-fidelity pdf sources with logistically significant size savings. Morphic assists in pre-RAG PDF prep for analysis, large-scale ingest & agentic analysis
Deterministic SEC 10-K item-level segmentation. Validated by structural invariants rather than hand-labelled ground truth. Zero API keys. Development record with Claude in prompts/.
Segment medical documents from smartphone photos using U²-Net, with tools for annotation, fine-tuning, and OCR-ready preprocessing.
To associate your repository with the document-segmentation topic, visit your repo's landing page and select "manage topics."