coaxial/ct2p repository overview

⁠combine-tif2pdf

This is a docker container whose sole purpose it to unpack an archive containing TIFF images, prep them for OCR using textcleaner, and combine them into a single PDF file.

Used in combination with coaxial/insaned-config, because a SBC isn't beefy enough to handle imagemagick on more than a couple of pages.

⁠Usage with Mayan EDMS

The container has two directories, dropbox, and processed. When a tar.gz file is written to dropbox, the container will unpack it, optimze each page for OCR with textcleaner, and combine them to a single PDF file.

⁠General flow

+-----------+     +-----------------+       +-------------+
|           |     |                 |       |             |
|  Scanner  +-----> combine-tif2pdf +-------> Mayan EDMS/ |
|           |     |                 |       | OCR         |
+-----------+     +-----------------+       +-------------+

⁠File flow between directories

+-----------+   +------------------------------+  +-------------+
| ./dropbox +---> unpack => textcleaner => pdf +--> ./processed |
+-----------+   +------------------------------+  +-------------+

⁠The dropbox directory

Is exported by the docker container at /ct2p/incoming, and should be mapped to a network share where scanned documents can be written to. The requirements for scanned documents are:

• Must be a tar.gz file
• Must contain .tif files

The pages will be assembled in alphabetical order, from their names. insaned-config names the pages with a timestamp, so they will be assembled in the order they're scanned.

⁠The processed directory

Is exported by the docker container at /ct2p/processed. Files placed there are PDF files, and can be readily ingested by Mayan EDMS. It can be automated by instructing Mayan EDMS to watch for changes in this directory and import any file written there.

⁠Limitations and TODOs

• It's more coupled than I'd like. Ideally, the scanner should do all the processing required for consumption by Mayan EDMS. Alas, the CHIP computer doesn't have enough RAM or CPU power to do this in any practical way.
• Handle files that aren't tar.gz by checking their extension and magic number. Files that don't match are deleted.
• Handle other formats than tif. Ideally, the script should find out what image files are in the archive and use this list for processing.
• Works best with 300dpi grayscale images.

⁠Credits

Uses GNU parallel⁠ by O. Tange, and textcleaner⁠ by Fred Weinhaus. For commercial/for-profit use, please refer to the textcleaner LICENSE⁠ file for more information.