Development of a handwriting recognition tool

• Status: Closed
• Prize: $1200
• Entries Received: 1
• Winner: arturneskromnyi

Winner

Contest Brief

In addition to recognizing handwriting in Portuguese, it must meet the following requirements:
- Accuracy of at least 95%.
- After object recognition (characters), the tool should search in the Portuguese language vocabulary for the most probable word for that set of characters. This should not be done for acronyms.
- It should allow the inclusion of new stretches of handwritting to improve (perfect/improve) training.
- The tool should allow administrators to include new words (from the technical field and that are not present in a common Portuguese dictionary) in their own dictionary that will also be used in handwriting recognition.
- The tool must recognize: paragraphs; segmentation of words at the end of the graphic line (translinearity); period; spelling accents; uppercase (capital); lowercase; subscript; superscript; and mathematical formulas.
- It should recognize the words, but not correct problems of accentuation and verbal agreement.
- The tool should not consider strikethrough text (scratched, erasure, Cross Out Text). For example: when someone misses a word, they make a strikethrough. In these cases, words should not consider.
- At the end of the recognition, you must export a .doc file (word) with the same name as the image file (photograph) used.
- All code must be commented out to allow changes by other developers in the future.

Note: The developer should evaluate whether in the inclusion of a very different or strange calligraphy if the overall performance is better by including an excerpt with this calligraphy for an individual training or if it could be included in the global training used in the recognition of all texts. If the first option is better, the tool should allow this possibility to create individual training for a specific person (student).

Recommended Skills

• Machine Learning (ML)
• Neural Networks
• OCR