During this webinar, industry experts discussed if computer vision computer is commonplace within enterprises that have machine learning models in production. We’ve included a short transcription of the webinar, beginning at 2:35 of the webinar.

Jesse Shanahan, Booz Allen Hamilton: To kick us off, is computer vision commonplace within enterprises that have machine learning models in production? Or is it still early days? If anyone wants to take this off, let's let's hear all about your experiences with enterprise-level computer vision.

Alyssa Simpson Rochwerger, Appen: We work with a range of enterprises, I'll just start things off, and I would say yes and no, depending on where that company is in the sophistication of their journey into machine learning. So some companies that we work with are incredibly sophisticated and have 10 year plus long-running computer vision models that do things like content moderation or even back office receipt transcription or check processing which are computer vision applications. Others are just getting started with their computer vision journeys and are much newer to it. So there’s a huge range, some companies completely rely on computer vision for their entire business and others don't.

Jesse Shanahan, Booz Allen Hamilton: I think that's largely been my experience as well. It really depends on the company and the organization. Some have deployed machine learning but it's very simple and very easy and explainable. But I found that with computer vision, you can strand that black box territory sometimes and that lack of explainability can be difficult in adopting it enterprise-wide. Has anybody else had similar experiences or different experiences?

Abhishek Singh, Apple: Yeah, that's largely true, what you guys said. Computer vision is often looked at as a niche field within the broader ML/AIcommunity. So I still see a lot of potential and I still feel that it's still early days for a lot of companies to adopt a lot of the technologies that come out of the computer vision field. And it's not just true for small companies. A lot of big players in the field that have established ML groups may not have fully realized the potential that could come out of using some computer vision techniques for their problems. 

Jesse Shanahan, Booz Allen Hamilton: Why do you think they haven't realized that yet? 

Abhishek Singh, Apple: It's probably because they've so far been busy trying to hit the bigger balls out of the park and focusing more on the traditional ML problems that can add more value to their business. But as those things saturate, the value adds can come from these other modalities, understanding images and videos better and see how that can improve their business model. So it's probably not as important for them yet. It could also be just because of lack of domain expertise in that area among people who are already there. So it's a bit of both that I see in a lot of companies today. 

Alyssa Simpson Rochwerger, Appen: Maybe I'll give a slightly different answer with two examples. So, one of them is a company that has been using computer vision for a really long time - the US Postal Service. They've been using OCR optical character recognition I think since about the 60s. They were probably one of the first companies to really put it to broad-scale, business-critical use and that's been going on for a long time. They have very sophisticated handwriting recognition models. On the other end of the spectrum today or yesterday or this morning, IBM actually announced they're getting out of the facial recognition business because they are underwhelmed with how the technology is being applied by some of the customers who are putting into practical use. 

I would say those two examples are good explanations of the range of what's going on in computer vision. So some companies have invested heavily and are continuing to invest and use this for mission-critical processing other companies are saying, ‘we did some experimentation in this field and it didn't go so well, we’re worried about bias that we're introducing, worried about unintended consequences, worried about negative applications of misusing computer vision technology so we're actually going to pull back from it. And so I think there's an incredible range just like there are lots of different technologies. I don't think it's unique to computer vision. It more speaks to the sophistication or the journey or the application that the company is on.

Jesse Shanahan, Booz Allen Hamilton: That makes a lot of sense. Daeil, were you going to say something? 

Daeil Kim, AI.Reverie: Oh, yeah. I was just going to add and confirm a little bit of what I'm hearing here as well. In terms of our particular experiences, we primarily work with Fortune 100 companies and what we're finding is that if it's not necessarily critical to the company's mission, then it's often sort of not nearly as invested. It's still early. I think it's really just early days. I think it's still pretty new. Unless you’re a company that needs to really get this done right.

I'll give an example of one company working with, Blue River, on helping them detect weeds in crops. For them, computer vision is critical. If you don't have a system that can actually detect that and segment that out with an algorithm, then you really don't have the technology that could be usable there. So that's where we really found a lot of excitement around but having said that, even though it's small, we think it's going to grow quite large once these tools are built. The technology is getting easier to use definitely and I would agree there. 

 

Learn more and watch the full video on YouTube: https://youtu.be/83QBypKQCP0.