Our vision is to mirror how humans understand visuals.

Visuals naturally group to define concepts. Labels don’t match to all visuals.

Fingerprints
are bio-inspired

A natural solution
for a complex challenge

  • Visuals come first
  • Are continuous & unlimited
  • Visuals group to concepts
OWL
MERGE
ANANAS

Labels
are man-made

A complex solution
for a simple path

  • Labels are an output
  • Are biased & supervised
  • Labels ungroup visuals

WHAT EVANGELISTS SAY

  • "To get neural networks to become intelligent on their own, as unsupervised learning, I suspect that means getting rid of back-propagation."
    Geoffrey Hinton - AI Pioneer and Google Researcher
  • "We clearly don't need all the labeled data. A real intelligence doesn’t break when you slightly change the problem."
    Geoffrey Hinton - AI Pioneer and Google Researcher
  • "Backprop is like Rolls Royce engine for airplanes. But they don't get us flying machines as agile & flexible as birds."
    Fei-Fei Li - Prof, Stanford; Co-Dir Human-Cntrd AI Inst.
  • "I'm deeply suspicious of back propagation. I don't think it's how the brain works. My view is throw it all away and start again."
    Geoffrey Hinton - AI Pioneer and Google Researcher
  • "Visual reasoning needs to be compositional. Without having seen a “person touching a bike”, the model should be able to understand it."
    Fei-Fei Li - Prof, Stanford; Co-Dir Human-Cntrd AI Inst.

We recognize visuals with limitless understanding and just one single sample to start.

null

VISUAL INTELLIGENCE

Ready – to – use
Computer Vision.
null

SINGLE-SAMPLE LEARNING

Autonomous AI with
Unsupervised Learning.
null

FINGERPRINT

Translating every image
into a unique fingerprint.

FINGERPRINT TECHNOLOGY

A Shazura fingerprint is a patented
mathematical representation.
Singular to every image or video.
Applicable to all visual content.

While Computer Vision models rely on annotating data…
Our Visual Intelligence works directly with fingerprints.

 

While training data delays the delivery of results…
Our unsupervised learning grants instant performance.

 

While machine learning is data-dependent…
Our fingerprints are unbiased and content-agnostic.

null
null

UNIQUE

Single descriptive
representation.
null
null

INSTANT

Real-time dynamic
generation & learning.

null
null

LEAN

Fast processing
& low storage.

null
null

SECURE

Natively encrypted
& consistent.
null
null

ROBUST

Strong persistence.
HW & content agnostic.

null
null

ACCURATE

+99% High-level
perception & recognition.

0
ms
INDEXING SPEED

Conversion time
per item to fingerprint

0
ms
REQUEST RESPONSE

Retrieval & search time
per 1:N fingerprint query

0
%
AVERAGE PRECISION

Fraction of accurate
recognized fingerprints

0
M
ROTATION & SCALABILITY

Ingestion & Indexing items
per day to fingerprints

FINGERPRINT AI CORE

How Shazura translates visual data into unique fingerprints to provide instant visual recognition.

1. Image/Video

  • Digitalized 2D/3D data
  • Computationally demanding
  • Non-descriptive representation
  • Redundant & noisy data

2. Shazura Patent

  • Unsupervised learning
  • High & low-level visual features
  • Top-down single translation
  • Optimized sparse–space

3. Fingerprint

  • Totally pixel independent
  • Natively invariant and lean
  • Low-dimensional depiction
  • Robust to acquisition conditions

FINGERPRINT AI LIFECYCLE

How Shazura fingerprints evolve from instant recognition to business use case customization.
null
1. VISUALS TO FINGERPRINT

Unsupervised learning.

Instant & ready-to-use.

Attention mechanism.

null
2. CURATED HYPERCLOUD

Data-driven curation.
Natural clusters.
Fast retrieval at scale.

null
3. NATURAL LEARNING

Natural propagation.

Use case business-centered.

Semi-supervised learning.