Handwriting parser - Githubissues

projectimacs / imacs

Visual Notetaking

0 stars 2 forks source link

Handwriting parser #17

Open francium opened 5 years ago

francium commented 5 years ago

Look into a simple but effective OCR solution.

Requirements

Able to handle PNG and JPEG
Minimal dependencies and setup
Usable on Linux, macOS and Windows
Produces good results on typical handwritten documents
Able to cleanly integrate with rest of application (need not be a node module, can be standalone application/library of a different language as long as previous requirements are still met)
Online APIs/SaaS discouraged

KelvinKKLin commented 5 years ago

Potential off-line libraries:

Tesseract.js (http://tesseract.projectnaptha.com/)
Ocrad.js (http://antimatter15.com/ocrad.js/demo.html)
GOCR.js (http://antimatter15.com/gocr.js/demo.html)
Ocracy (https://github.com/naptha/ocracy)
Node-tesseract (https://github.com/desmondmorris/node-tesseract)
Node-tesseract-native (https://github.com/mdelete/node-tesseract-native)

See this link for a list of other OCR libraries: https://github.com/kba/awesome-ocr:

francium commented 5 years ago

I know @khand5 want's to move forward with MyScript, but I would like to continue research into offline solutions as well while there is time.

KelvinKKLin commented 5 years ago

I want to test Tesseract and Ocrad to see if they're any good. If so, I'll make a demo for it.

KelvinKKLin commented 5 years ago

Update: Tesseract, Ocrad, and GOCR are good with pictures/scanned images of books and printed text, but not good with handwriting.

Update 2: Apparently, Tesseract can be trained to recognize handwriting, but I haven't investigated this in detail. Don't know how much effort this will take.

khand5 commented 5 years ago

Wrote a second handwriting (non-MyScript) solution in #57 that meets all requirements above except for handling images.

@francium Please review #57.