Handwritten Character Recognition

Gave a talk on handwritten character recognition at RAIT, Nerul workshop on pattern recognition today. Loosely derived from my student Leena Ragha’s PhD work, the talk attempted to look at broader space of HCR.
Questions/issues raised during the talk include:

  • how do humans recognise handwriting – what do we see: characters, word as a whole, some other pattern? We are not particularly good at handwriting recogntiion in general; but we do manage well if it is a document (words/sentences/etc) thanks to context.
  • what is the gold standard? Given the not-so-great human performance, what are we comparing against. We are doing just a percentage recognition game with various corpus. My 91 is better than your 90! Will these results port across corpus? across languages?
  • what exactly is the learning engine learning? The learned knowledge is rarely visible thanks to use of techniques like NN and SVM. We are being too obsessed with recognition accuracy.
  • etc.

Humour point: Google is like Hanuman! When asked for the plant of mritasanjeevani Hanuman bought the entire mountain containing that plant. Just like google when you ask a query, brings up a million pages most likely containing the answer to your query somewhere.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: