Extractor MVP

This MVP extracts chart data, tables, text, and image captions from PDFs, images, and DOCX files into structured JSON and XLSX outputs.

Features

• Chart digitization using the free google/deplot model (image-to-text) for graph value extraction.
• Image captioning using Salesforce/blip-image-captioning-base.
• OCR for text extraction using easyocr (no Tesseract).
• Table extraction from PDFs via camelot with multi-page table merging.
• Table/text extraction from DOCX via python-docx.

Usage

python -m extractor_app.main /path/to/file.pdf --output-dir output

Output

• output/extraction.json with structured data.
• output/extraction.xlsx with separate sheets for charts, tables, text, and images.

Notes

• For best results, provide high-resolution images and vector PDFs.
• The MVP focuses on extraction. You can extend post-processing to map axes, detect units, or normalize dates.