While creating the entry for Fast Deep/Recurrent Nets for AGI Vision, I took particular note of the unbroken hand writing competitions. That task, for computer vision, is more difficult than “segmented” hand writing with breaks between the letters.
Are there parallels to subject recognition as performed by our computers versus ourselves?
That is we record and use “discrete” values in computers that are used for subject recognition.
We as human observers report “discrete” values when asked about subject recognition but in fact recognize subjects along non-discrete continuum of values.
I am interested in the application of techniques similar to continuous handwriting recognition applied to subject recognition.
Comments?