Skip to content

DocEdit-PDF release that includes user request, image, and command for each edit.

Latest
Compare
Choose a tag to compare
@vmorariu vmorariu released this 18 Apr 17:06

This is a release of the DocEdit-PDF train/val/test splits that includes the user request, image, and command for each edit (the test set currently does not include the command).

The train/val portion of the release has been split into four archives so that each fits within the file size limit. Please extract all archives in the same directory.

The Enron Email Data and government forms are in the public domain or are licensed under their respective licenses (as applicable). All other materials, including the modifications made to the original documents (if provided) and the descriptions of the modifications which include the natural language description and the structured command-like descriptions that include the bounding boxes, etc., are licensed under the Adobe Research DocEdit Dataset License, Version 1.0.