Inspiration vs Reality

Extended Visual Pathway (EViP.1)

ADC’s image processing system is based on the mammalian Extended Visual Pathway, incorporating features like a saccadic eye movement emulator, a bio-inspired visual pathway filter array, and a linear visual cortex to enable the detection and recognition of objects in open and ambiguous environments. This is the key to EViP’s ability to effectively mimic human visual capabilities, particularly our ability to recognize an object or person in a single glance, even when only seeing a partial, low resolution, or sketch image.

EViP vs HViS (Human Visual System)

For a single-sample database, detailed testing has shown that ADC’s EViP system can match human visual system performance (75% correct) in a face recognition task involving a database with 1,000 distractors. With 10,000 distractors, EViP outperforms HViS, providing 66% versus 24% correct recognition in rank 1.

Table I: HViS vs EViP comparison based on MegaFace Challenge of UoW

Distractors	1K	10K
HViS*	75% Correct/ Rank 1	24% /Rank 1 & 91.1% /Rank 10
EViP	75% Correct/ Rank 1	66% /Rank 1 & 78.5% /Rank 10

* This results are from   http://megaface.cs.washington.edu/results/humanstudy.html

HViS vs EViP Comparison

Blue-HViS
Pink-EViP

Intelligent Perception and Cognition (IPCO)

EViP’s real-time feature extraction engine-unsupervised learning technique provides sufficient object features for full/partial, noisy and low resolution, incomplete and sketch detection, recognition, and identification and facilitates autonomous adaptive capability in short term memory (Retina block) while EViP’s non-competitive supervised learning technique enables autonomy, in situ, adaptation learning to provide cognitive capability in long term memory (Visual Cortex block) for intelligent perception.