Thursday, September 2, 2010

Vision as a Service?

They say that if you see a bandwagon rolling you’ve already missed it, but in this case I hope I’m in time to catch a ride.

The bandwagon I’m chasing is “Software as a Service,” usually described by the acronym SaaS. SaaS means, at least as I understand it, that the user never actually buys software, he or she just pays for the ability to use it over the internet, presumably on either a per hour or fixed fee basis.

So here’s my idea: machine vision software is a real pain to develop, maintain, upgrade and so on, so why don’t users just pay for it as a service?

Let’s think this through. Suppose a big machine vision company decided that rather than selling software it would provide software over the web to industrial users. At a stroke they eliminate all their worries about unauthorized copying of software, plus they can release patches and updates quietly in the background. Users meanwhile would see little change in terms of the interface they use. They would continue to develop their applications on a PC and roll them out to the plant floor as needed.

So what’s the advantage?

Well, assuming a networked factory, (yes, I know that’s still something of a stretch,) the application could be accessed from anywhere. That means once the initial image acquisition has been engineered the bulk of the application – training new parts, managing inspection statistics and of course, algorithm development – could be done anywhere. In effect, a manufacturing company could subcontract the management of their machine vision systems.

Other benefits include the security of off-site data backup, plus no need to invest training in people who then go to work for competitors, no getting stuck with old software versions …. The list goes on.

Why, if this worked out, vision applications might be developed and supported from India at a fraction of what we pay to do it here in the US …

Scratch the idea, it’ll never work. I’m talking nonsense. Keep buying machine vision software and hiring experts to work on it in your facility. Please.

1 comment:

Jakob Kirkegaard said...

It is a great idea - I have been pondering the same. I guees the biggest obstacle is latency etc in transferring large images to the processing servers in a production process requiring real time response.