Open Source Handwriting Recognition
summary
An algorithmic, non-ML approach to handwriting recognition using Procrustes analysis for curve matching against stored shapes. Implemented as a proof of concept for Malayalam and Tamil, with potential extension to other Indian languages and use as a privacy-preserving mobile typing tool. The approach emphasizes open-source, data-light recognition rather than large proprietary ML models.
AI-written summary · about
chapters (10)
- 0:00#Indias
- 0:27GOOGLE KEYBOARD WORKS FINE, RIGHT?
- 0:57PRIVACY KEYBOARD IS IMPORTANT NO INTERNET PERMISSION FOSS REASONS
- 1:46HOW IT WORKS? CURVE MATCHING
- 3:01PROCRUSTES ANALYSIS SCALING ROTATION DISTORTIONS TOLERANCE FOR ERRORS
- 7:04OVERLAPPING LIGATURES അനീഷ് ആന്റി
- 8:06TRAINING UI COLLABORATION FROM NON DEVELOPERS
- 8:31DICTIONARY & MARKOV CHAIN BASED PREDICTION മുഖ്യമ - മുഖ്യമന്ത്രി മുഖ്യമന്ത്രിയുടെ വാർത്താസമ്മേളനം
- 8:41DISAMBIGUATION പഠനം മഠം 101
- 13:11MORE DEVELOPERS MORE LANGUAGE EXPERTS
related talks
40mBuilding a Handwriting Recognition System
44mBuilding Handwriting Recognition System for the NY Times Crossword
06mWrite, Attend and Spell: Streaming End-to-end Free-style Handwriting Recognition Using Smartwatches
08mIndian Sign Language Recognition (ISLAR)
27mSign Language Recognition Improvements Since the London Keynote
44mHello Deep Learning
25mBuilding a Handwriting Recognition System for the New York Times Crossword
27mIndian Sign Language Recognition(ISLAR)