v1.2.1

Release 1.2.1

Release 1.2.0 – bug fixes and PDF support

• Two bugs related to running with multiple threads (the default) are hopefully fixed in this version:
  • Sometimes the annotated image generated from a given service would have the results from another service also written over it. This was inconsistent and intermittent, and the exact cause is still unclear, but this version of Handprint implements a workaround that hopefully stops this from happening.
  • The Amazon service interface would sometimes fail with an error related to endpoint_resolver. This was due to the previous code not following recommendations about using a different connection session object in each separate thread. This is now fixed.
• This version changes the way output files are named. The new scheme always includes the string handprint in the file name for easier recognition and to help reduce the chances of file name collisions. The scheme uses the naming pattern somefile.handprint.png for
  the rescaled input image, somefile.handprint-service.ext for the
  various service output results, and somefile.handprint-all.png for the
  summary grid image. (Issue #10.)
• PDF files are now accepted as input. (Issue #11.) Note, however, that
  Handprint will only extract the first image in the PDF file.
• The order in which resizing and rescaling is done has been swapped: if a file is too large, Handprint will first rescale it, and then if it is still too big (in byte size), it will resize the file. This appears to result in images that have higher resolution than the previous approach, which did the steps in the opposite order.
• The URLs in the file of example URLs, tests/urls/caltech-archives-urls.txt, have become invalid. They have been replaced with other URLs that are valid (as of right now, anyway).
• A few more bugs have been fixed.

The file CHANGES contains a more complete change log that includes information about previous releases.

Release 1.1.0 – Support for comparing results to ground truth

This release adds support for comparing extracted text to expected (ground truth) text. To use it, users need to provide text files that contain the expected text for each input image; Handprint then produces tab-separated files listing the errors (compared to ground truth) in the results received from HTR services. Please see the relevant section in the top-level README file or the application help (via -h) for more details.

Other changes:

• The debug option -@ now accepts an argument for where to send the debug output trace; the behavior change of -@ is not backward compatible.
• The installation instructions are (hopefully) improved and simplified.
• There are internal architectural and organizational changes.
• There are some bug fixes.
• Most test images have been moved to a separate repository on GitHub.

The file CHANGES contains a more complete change log, and includes information about previous releases.

Release 1.0.3 – Bugfix release

This version fixes an internal bug involving the credentials file used for Amazon services.

Release 1.0.2 – Bugfix release

This release fixes issue #9, a problem that credentials files were not being saved in the expected locations.

Release 1.0.1 – minor bugfix & addition of info

This version fixes a bug writing files downloaded from URLs and adds instructions for installing from PyPI.

General information: Handprint is a Python program to apply different handwritten text recognition services to images of handwritten text pages, and produce an annotated image (and optionally more) showing the text recognized. More information can be found at the project home page. The file CHANGES contains a more complete change log that includes information about previous releases.

Release 1.0.0

This is the first sufficiently complete and robust version of Handprint that can finally be called a version 1.0. There are very many changes compared to version 0.10.0, including a new way to provide credential files, new default output (in which results from different services are placed side-by-side in a single large image), new command-line arguments, parallel execution, and more.

The file CHANGES contains a more complete change log that includes information about previous releases.

Version 0.10.0 – Support Amazon Rekognition and Textract

This release primarily adds support for Amazon's Rekognition and Textract services. It also updates some internal code and fixes some small bugs, and adds a code of conduct and contributor guidelines to the repository.

Version 0.8.1 – bug fixes

Detect and handle when the Google API returns a badly-formed bounding box. Also skip files previously generated by the annotation feature of Handprint.

Version 0.8.0 – Generate annotated images

The latest version of Handprint will create copies of the images in files named with the method extension (e.g., document-1.microsoft.jpg) and write overlays of the text and bounding boxes extracted by the methods. This makes it possible to see the text extracted directly over the source image. Generating these annotated images takes a little bit of time and if they are not necessary for your purposes, you can turn off annotation with the -A option (/A on Windows).