deepdoctection - An open source package for document intelligence
summary
An open source pipeline called deepdoctection assembles layout detectors, table segmenters, OCR wrappers (includes Tesseract and pdfplumber), and matching/reading‑order steps so documents turn into NLP‑ready structures. Pretrained models on PMC papers reach about 0.9 mAP on layout detection, fall to ~0.5 on 500 hand‑labeled business reports, and recover to ~0.85 after ~600 pages of targeted fine‑tuning. Code, training scripts, and notebooks are provided; models live on Hugging Face.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 0:40Janis Meyer
- 2:00Why deepdoctection
- 3:23The problem
- 4:22The solution
- 8:31Example
- 11:00Extraction
- 13:06External packages
- 14:50Data sets
- 16:32Training script
- 19:42parsing
- 20:04pipeline
- 22:03output structure
- 22:37overview
- 23:09ideas
- 24:49questions
related talks
01h 19mLearn to Unlock Document Intelligence with Open-Source AI
01h 25mLearn to Unlock Document Intelligence with Open-Source AI
19mDocument Reader using Deep Learning with a Small Dataset
42mUsing Deep Learning to Understand Documents
01h 00mComplex Data Ingestion with Open Source AI
29mConquering PDFs- document understanding beyond plain text
44mExtracting Structured Data from Legal Documents
40mDeep Learning and Optical Character Recognition