Decapod Accessibility

• Use proper AIRA status: visible/hidden off screen

• Keep a consistent interface regardless of accessibility, even though collapsing panes is primarily for visual users. (Non-visual users may need help from visual users, therefore both interfaces should be the same).

• Quality of an image can be discerned by examining the resulting OCR'ed text.

• A spellchecker can be used to grade accuracy.

• To compare OCR across two images, compare the spell check results of two images.

• Once properly set up, capturing shouldn't be a problem. A sighted user can help if needed.

Capture > OCR > Spell check > give accuracy grade.

In order to achieve this kind of functionality, Decapod should have the option of doing OCR to images as they are captured for QA purposes for screen readers. Speed of OCR will be a factor.

Screen Reader Workflow 1:

• Use Decapod to do capturing to image

• Then put those images into Kurzweil 1000 or OpenBook on Windows for scan and recognition.

Workflow 2:

• Use Decapod to do capturing to image

• Let Decapod do batch scan and recognition

• PDF is fine as long as it is tagged properly (paragraphs, headers, etc.)

• HTML is good as long as it is reasonably well marked.

• Should be open and extensible so that users can add / extend functionality (i.e. add scanner support, etc.).

• When creating a Live CD, be sure to include the gnome-accessibility meta package

• ORCA Mailing list

• Ubuntu Accessibility

• knfb Mobile Reader

A separate project can be created for generating text from images

• What is required is a single camera (monocular), extracting text lines, and OCR

• no need for heavy dewarping, and other visual processing.