In natural learning, humans are capable of recognising variations of the same character and reproducing the character in different sizes and presentation styles. This imprint or ‘concept’ of a learned character may be influenced by new attributes while maintaining its general structure. Hence, in line with natural learning, the image recognition system must return the closest matching result, and also, secondary matches with the high resemblances to the closest matching result.

https://neuramatix.com/projects/videos/DigitLearning.mp4

Character recognition powered by NeuraBASE.

https://neuramatix.com/projects/videos/DigitLearning_IBM.mp4

A visual recognition task performed by a neuromorphic chip built by SyNAPSE researchers at IBM.

Source: https://youtu.be/oKvatA-ec4k