Skip to content

Latest commit

 

History

History
36 lines (24 loc) · 1.95 KB

README.md

File metadata and controls

36 lines (24 loc) · 1.95 KB
Raw

Overview

This directory contains source code for the OpenMPF Tika image detection component.

This component extracts images embedded in document formats, such as PDF (.pdf), PowerPoint (.pptx), Word (.docx), OpenDocument Presentation (.odp), and OpenDocument Text (.odt) documents.

In general, the Tika parsers will extract unique images, once per track. Each track will contain a PAGE_NUM property specifying where the image is embedded. Note that page numbers start at 1, not 0, unless otherwise noted.

The path where the extracted image is stored will be reported in the DERIVATIVE_MEDIA_TEMP_PATH property for each track.

By default, extracted images are stored in $MPF_HOME/share/tmp/derivative-media/<job#>/<uuid>/tika-extracted. Users can set ORGANIZE_BY_PAGE to true to store each image in a sub-directory labeled by page number (for example, page-1) . Images that appear on more than one page will be placed in a common directory instead.

Note that the OpenMPF Workflow Manager will move those files to a more persistent storage location reported in the JSON output object for each piece of derivative media with the job is complete.

Format-Specific Behaviors

The following format-specific behaviors were observed using Tika 1.28.1 on Ubuntu 20.04:

  • For PDF files, the first page corresponds to a PAGE_NUM value of 1. Multiple pages can be reported in a PDF document and are separated by semicolons. For example, PAGE_NUM = 1; 2; 4 , would indicate the embedded image appears on pages 1, 2, and 4 of a PDF document.

  • For non-PDF files we intentionally set PAGE_NUM = -1 to indicate that the page number cannot be determined.

  • OpenDocument Presentation documents will generate a thumbnail / preview .png of the content of the last modified slide, including text, even if it's blank.

  • OpenDocument Text documents will generate a thumbnail / preview .png of the content of the first page, including text, even if it's blank.