Ground Truthing tool for Color Images with Text

In the research of algorithms that extract text from color images a set of files with the exact location of the text is needed to avoid inefficient and tedious visual checks of the results.

This Ground Truth information saves enormous time and gives accuracy.

For this the gttext project helps to create fast and quality Ground Truthed data-sets from color text images.

The basic level of work is at a pixel detection, making possible to group regions to form the glyph or even use a direct editing to get the choice.

This computer engineering project was the result of a degree requirement and it is offered to the open source community.

Under GNU License, this project is open to improvements and to accept developers that may want to give a positive change.

Enjoy.

Update 1.1.2

• Faster performance
• New mode of region selection
• Higher scalability for up to 70MP
• Compressed Ground Truth data definition

Information stored as runs (x,y,r) instead than single pixels (x,y) definition_sample

Update 1.2.1

• Improved pixel edition tool

Update 1.2.2

More details at Versions List

Developer Update 1.2.6

Release 1.3.0

• Fastest selection
• Fixed Bugs
• Image Exporting

- With the help of the OCR Tesseract project from HP and Goolge,

• Text extraction from the selected pixels.

The recognition of text copies the selection to clipboard with very high accuracy in greater than 10px/letter ratio aspects images.

- Added option to select from all the image.

• Full text selection to obtain all the regions similar to the one clicked.

Detect those regions whose color changes more abruptly makes it an ideal tool for readable text capture.

Release 1.3.4

• Fixed important Bugs in Windows 7
• Fixed Bugs with main icon
• Reduced greatly the installer size

I want to thank all of you using the software for your confidence in this project.

Version 1.4.1.

• Fast creation of new projects from the clipboard image content.
• More Keys shortcuts to execute faster the most common operations.
• Inner framework updated.
• Bug fixes.
• Improved OCR detection.
• Select the area to perform text extraction.

Less clicks, more stuff.
Enjoy
David

Version 1.4.3

• New multi-language OCR GT Text

• Better compatibility with old Windows XP systems

• New tools to extract text from selected pixels

Release 1.4.4

• Importation of documents or images from Scanner or Camera

• Short cut Ctrl+V to copy fast a new image document from the clipboard

All in all, just takes seconds to have the text from any image in any circumstance.

Enjoy, David

:)

FileBuzz.com has agreed to host a link to this project and has awarded it with 5 stars

Thank you!

Thank you very much also to