Sightech Machine Vision - Self Learning Eyebot
Self-Learning PC-Eyebot
Sightech Home Page
About Sightech
Contact Sightech
Frequently Asked Questions
News, Events, and Shows
Dealers and Resellers
Vision Products PageApplications PageSupport, Documentation, etc.Intelligent Vision ExplainedSightech Careersfiller

Reasons for Choosing Trainable Vision

Although traditional machine vision has evolved over the past decade, it still suffers from the limitations of its general approach. Frequently, real world problems need a vision system that can process the enormous complexity often presented by real world problems. The approach of old-world vision solutions utilize an ever-growing repertoire of tools which are then used in an attempt to deliberately and manually simplify the real world to the point that it can be measured and judged. All too often this approach ultimately produces partial results, resulting in a painful situation where the user is forced to compromise his original inspection goals. This is not the purpose of machine vision.

Items to consider: Legacy Machine Vision: Today's Trainable Machine Vision:
Initial demonstration for qualification of approach? No! Usually cannot be fully demonstrated before installation.

Yes! Usually can be fully demonstrated in a single visit in a matter of hours.

Trial setup?

No! Usually not feasible - full installation is often required for effective trial

Yes! Due to the benefits of trainability, trial setups can be easily accomplished.

Cost, time, and effort of setup?

High cost: Requires elaborate tool deployment and/or programming. This work is entirely dedicated to the particular inspection task.

Minimal cost: Reduced setup effort because trainable vision does much of the work itself..


Can vision perform original inspection task goals?
No! Due to the vast complexity of real-world inspection problems, the initial inspection goals are often compromised. Yes! Trainable vision often surprises users because it can process the massive complexity of everyday vision problems and deliver to original expectations.
Pass/Fail or Absence/Presence?

After work: Many criteria must be deliberately combined to yield what most problems need in the end - a "Go/No-Go" decision.

Built-in: The primary and natural result of trainable vision is "Go/No-Go". This is accomplished by the massive internal work performed automatically by trainable vision systems.
Process or product changes?

Difficult! The original expert personnel often need to come back and make changes for even the slightest shifts in process or product. Ongoing support can be difficult to impossible. This is why a lot of these systems are eventually taken off the line.
Easy! Trainable vision allows the users to easily perform brief cumulative training that automatically learns and accumulates the new knowledge about the process shift or product change. This can be implemented as a single "Learn" button placed on the line and set up to learn for a preset period each time the operator pushes it. Internal backup makes going back easy if necessary.
Vision of the future?

No! The old tool-based and programmable approach is giving way to modern vision systems with built in intelligence. Users are starting to demand more.


Yes! Trainable vision employs state-of-the-art artificial intelligence technology which allows the vision system to inspect products much the same as persons do - intuitively. The trainable vision system automatically develops the necessary familiarity needed to perform successful inspection.
Can it actually do the task?

Often No! Many difficult applications cannot be solved with traditional machine vision at all. Subtle defects complicated with reflections, etc. often make detection impossible.

Often Magical! Because of the sophisocated self-learning algorithms internally employed by trainable vision, previously unsolvable vision problems can often be easily tackled. This is when trainable vision's performance is so much better, it seems magical.
New groundbreaking abilities?

No! Traditional machine vision uses old techniques such as measuring, template-matching, pixel counting, optical character recognition, barcode reading, etc.which are non-trainable offerings in comparison with trainable vision.

New Modes! Sightech is offering new modes that take machine vision, trainable or not, into brand new areas of image processing. One example is Coloration - a brand new mode that takes color detection from a spectral understanding to a 'Shape of Color' understanding. This important mode easily solves color detection problems that were previously impossible. You can get Coloration only from Sightech.


Home | About Sightech | Contact | Products | Applications | Support | Technology | Careers | Site Map | Resources

Copyright © 2005 Sightech Corporation — All rights reserved
Privacy Statement

Valid XHTML 1.0!

Valid CSS!