Support PDFs as input type

eamonnbell-dur / maoii-workflow

Workflow for massive analysis of internet images

MIT License

0 stars 0 forks source link

Open eamonnbell-dur opened 9 months ago

eamonnbell-dur commented 9 months ago

In order to do this we need to extract images from the PDFs, which can be done in the following ways:

Direct extraction of any images embedded in PDFs (e.g. pdfimages)
Rasterisation of PDFs to page images (and then)
- Document layout analysis and filter by "image" classes
- First pass of SAM model (on page images)