Episode 34
Fabio Perelli
Zebra Technologies

 

On this episode of Manufacturing Matters, Fabio Perelli, Senior Manager, Product Management at Zebra Technologies joined Winn Hardin and Jimmy Carroll at the A3 Business Forum 2024 to discuss some of the most exciting technologies and applications within Zebra’s machine vision and fixed industrial scanner group. Topics included the latest developments in deep learning and 3D imaging in areas including electric vehicle battery manufacturing, warehousing and logistics, and packaging. Additional topics including cloud computing, the Zebra acquisition of Matrox, and what’s in store for the immediate future.

zebra technologies logo

Zebra-Fabio-Business Forum 2024-audio PG.m4a: Audio automatically transcribed by Sonix

Zebra-Fabio-Business Forum 2024-audio PG.m4a: this m4a audio file was automatically transcribed by Sonix with the best speech-to-text algorithms. This transcript may contain errors.

Jimmy Carroll:
Hi, everybody. My name is Jimmy Carroll. I'm the vice president of operations at Tech B2B Marketing. I'm here for the Manufacturing Matters podcast. I have the pleasure of being joined by my colleague Winn Hardin and Fabio Pirelli from Zebra. Fabio, thank you so much for taking the time. It's always a pleasure to see you. For those who don't know, if you could give us a quick background on Zebra and what you do there, I think that would be helpful.

Fabio Perelli:
Sure. Okay, so I'm a product manager. I work in the product team. I have a team of guys that help me out, and basically we manage all of the hardware development on the machine vision side of things. And you also probably know my counterpart Pierantonio. He also looks at all the software, because today if you really look at the products, you have a combination always of software and hardware. A piece of hardware on its own, it doesn't really do that much. So you need to have this combination. So Pierantonio and I co-jointly develop and follow up development of all the machine vision products there.

Winn Hardin:
Now Zebra does a lot of other things too in addition to machine vision.

Fabio Perelli:
Correct. So we are part of the Machine Vision and Fixed Industrial Scanner Group. Zebra has a division that does printers. Then you know the end health scanners, RFID, many, many other products, but the group I'm in is machine vision–centric.

Winn Hardin:
Well that's good. You've done that cross-pollination between the different product lines. It makes you stronger as a company in terms of your offerings for sure.

Fabio Perelli:
Well, it opens doors. You sell RFID and well, maybe you want to inspect your products.

Jimmy Carroll:
I want to ask about that. And I'm going to ask about some specific technologies within the Machine Vision Group. But before I do, I have to ask, how has the transition from Matrox to Zebra gone and what's most exciting about it to you?

Fabio Perelli:
The transition is I would say pretty much completed. There's always a little hiccup here and there, but overall I think everybody did a great job. It was a pretty big undertaking that took quite a bit of time, but I think it worked out. We're pretty much a cohesive team at the moment, so it's very, very good. One of the things that got us very excited when we got the news that we were joining Zebra was the fact that we had very complementary product lines. And I'm talking mostly, again, in the machine vision business units. They had a smart camera, yes. And they had industrial scanners, yes. We had a smart camera too, but it was at a somewhat different level. And the software offering was very different. On the other hand, today we have lots of software to offer, which is a great advantage because if one package doesn't do something, we have something else to offer as a complementary product, which is nice. Library tools — we've got mountains and mountains of functionality there, so I think we're in a pretty good position.

Winn Hardin:
Back in the day, machine vision systems would almost be standalone nodes, but that's not the way it is now with IoT, Industry 4.0. And that's not just buzzwords anymore. I mean, we're really seeing that communication across, and not just in the plant now, but as the cloud, and I would love to hear what you guys think about the cloud. And there's a lot of discussion about whether this is going to be the year of the cloud. The PLCs, the other programs are going to start having that embedded in, it being more of a requirement. I've always been a bit of a skeptic, because manufacturers want to protect their data at all costs. So it's a hard, difficult hurdle to overcome a little bit from acceptance.

Fabio Perelli:
Yeah, the cloud has been a lot of talk for, as you said, many years. A lot of our customers are still somewhat nervous about it because their data, which is their secret, is transiting who knows where. On the other hand, what we have seen for the deep learning or AI stuff, you bring the cloud on-site. So you actually set up a computer farm right on your facility. So you get all the benefit of the big computing, but then you keep everything local and some other ones you're able to encrypt data. OPC UA is a good example, where everything is encrypted, A to Z, for the protocol, hardware, everything is encrypted. You should not be as concerned in my opinion. So yeah, the facilities could be used without too much worrying about losing or competitors taking bits and pieces of your secret sauce there.

Jimmy Carroll:
AI is something I wanted to talk about. So Zebra, there's a lot to talk about in the Machine Vision Group, from barcode scanning and machine vision cameras and smart cameras. And we could talk about this for hours, but one of the topics I want to ask about is deep learning. So it's a buzzword I understand. But in the last couple of years, it's really seemed to become a little bit more of a mature technology. And as people start to realize that it's a tool for the machine vision toolbox, it's not a magic tool that's going to solve everything, but there are very useful, practical applications for it. And one of those is OCR. So your tool is designed around OCR, designed for OCR. How can this make the vision and the vision engineer or the systems integrator's job easier?

Fabio Perelli:
Yes. A very good question. I mean, overall, the deep learning had been more of a buzzword, a lot of science projects, PhD thesis. I think now we're at the moment where we can actually physically use it in practical applications to solve real problems. What you're discussing here is what we call the deep learning–based OCR tool. It's pretty powerful. I mean, I've been in the industry for 27-odd years — unfortunately you can you tell my age. On the other hand, I did see a lot of these issues — reading, texts, and all that stuff. This deep learning–based OCR tool, it really freaked me out because it works. And it works out of the box. There's no complex training, which with a lot of the previous versions of OCR tools, you had to get the alphabet out explicitly, say this is an A, a B, a C, and it takes forever. And then you get a slight variation and then you have to start over again. This thing, you just get an image with text in it, and it reads it, irrelevant of the size, the shape. It was pretty impressive.

Winn Hardin:
And for me, deploying deep learning, one of the biggest hurdles has been getting the integrators. Because the integrators got burned on a lot of the early — I mean, forget about the neural network stage, but when AI/deep learning was first coming out, a lot of people spent a lot of time trying to solve a lot of applications with it and were not always as successful as they wanted, to see that ROI. So I would think that if you're able to use it in a finite space, like in an OCR application, and it can continue to learn and evolve in a continuous learning environment, that seems like maybe it'll start to create that confidence as well as the familiarity.

Fabio Perelli:
Again because training a network is not that straightforward. So we give you something that's pre-trained, with potentially millions of images that we used to get to that point. So obviously when you show something it will probably work right out of the box. So then therefore the integrator has very little time to waste trying to get everything to work. He's almost guaranteed that it's going to work out of the box without any crazy lighting setup.

Winn Hardin:
And then three to six months putting a temporary system on the line so you can acquire the datasets.

Fabio Perelli:
Exactly. So you bypass and jump over all that problematic stage, which is very time-consuming. Costs money. To have something that you can solution and show the customer: This is how it works and it works quite well. At Automate, I don't know if you remember, we had a demo with a rotating table that had four different stations of four things that were totally different. One was like metallic dot pin text. It reads it, with the same lighting as when you're trying to read something on a curved bottle. And there was something and there was like inverse contrastive . . .

Winn Hardin:
We're talking nonreflective surfaces in that particular case, which is really astounding.

Fabio Perelli:
Embossed, I mean, it really, really, really solutioned the problem without having to create special lighting for each one of these scenarios. Again, I was very impressed with that because typically you need to do a lot more work on that.

Winn Hardin:
You're building that where it combines photometric with standard two dimensional, and I think three dimensional is something else you wanted to talk about today too. I mean, does it come with a standard optical head?

Fabio Perelli:
No, typically, depending on what product you're going to choose. If you choose, for example, the RS-GTX, you're free to pick your own lenses and your own lighting so you can set up whatever distance you want it to work, whatever lighting conditions and all that. We also have a line of what we call the VS cameras, which came from the Zebra side. Well, some of them have the complete illumination built in. You have liquid lenses for focusing. So a more complete package. It works, maybe not in every condition, but yes, it's easier to set up for an integrator, for example. And the algorithm can run in either one of these platforms. So all of a sudden you have the flexibility of choosing very different and variable forms with software that's kind of common, so you know that's going to work, irrelevant if you pick A or B, because it all comes from the same . . .

Winn Hardin:
And that's the true power . . . I mean, we've rarely seen enough of that in machine vision, where a software can be truly hardware agnostic. I mean, it's usually going to be very much a paired system between the software, because machine vision is difficult to be successful at. But it's really cool to see software evolving, whether it's an IDE with an AI OCR tool built into it, to basically say: You can run on any of our platforms. But frankly, if you're a Ford or GM and you're married to certain other hardware platforms, our software can still be the powerhouse that makes you successful.

Fabio Perelli:
A lot of flexibility.

Jimmy Carroll:
I think it speaks to the whole idea of ease of use. That's one of the terms that keeps emerging. And maybe it's an obvious one. Everything needs to become easier to use. But do you see that as a main growth driver in the increased adoption of automation? for example, obviously this is a product along with many of your others that can be deployed in warehouses. So warehouse automation market intelligence company Interact Analysis predicts modest growth for warehouse automation in 2024 but double-digit growth in the years to follow. Do you see ease of use as one of the main growth drivers there? And what else do you see?

Fabio Perelli:
I think ease of use is definitely one of the key things that we are also working on. We're trying to get that learning curve of learning a software or understanding how things work as simple as possible. Because that's time wasted, because people want to solution things. They don't want to learn. People know the vision. They understand vision. They don't need to understand how our package works. They just want to use it in the best possible way to get a solution up and running. So this is definitely one of the focuses that we have. And, pardon the pun, but the Aurora Focus, which is what Zebra has developed for the smart camera line and for the fixed industrial scanners, it's very simple to use. You can get up and running very quickly, and we're going to evolve this more and more. We're going to continue developing that. And we also have the Aurora Design Assistant, which comes from the Matrox side of the business, where if you need more complex, more evolved type of functionality, you have that as well.

Winn Hardin:
That's really what I wanted to ask, because you can have super-powerful algorithms, quote unquote idiot-proof algorithms. But it takes more. When you talk about ease of use, I was going to push you on that Fabio a little bit, just because everybody says our software is easy to use, but they don't really talk about illustrative examples, all the different little parts. So Design Assistant is a critical element, as well as the powerful algorithms behind it. Are there other aspects in terms of, and I know it's like talking about machine vision and then talking about training, you rarely do that. But in terms of knowledge bases, in terms of in-line help systems, which the Design Assistant provides part of that. But are there other things that Zebra does to truly make that easy-to-use software? Is it primarily the algorithms and the Design Assistant or is it more of a whole, also, with what you've got on your website, you have educational programs?

Fabio Perelli:
We do have a portion of the website that has training, online training. So you can self-train. We offer physical training courses. We offer online courses. So there's a whole slew of ways to get familiarized with how to use the software. We're looking at how to leverage some of the ChatGTP type of artificial intelligence to cater to . . .

Winn Hardin:
We might as well eat our own cooking.

Fabio Perelli:
So something that we're looking into is an easy way to — you ask a question in plain English and you get . . .

Winn Hardin:
the answer that you truly need.

Fabio Perelli:
Correct.

Winn Hardin:
And that's one of the real benefits of working with one of the larger companies, especially machine vision, which is traditionally been a lot of little companies with a couple of monsters in there. You have the resources to be able to offer all these systems. So the training isn't an afterthought, so that aftermarket support is there without always having to push back to the integrator. Integrators don't always prioritize those maintenance contracts long term, which is always something over the years that has surprised me. But I guess we can just make more revenue. But I wonder, as far as when you're designing and solving an application, that's obviously super-critical, but are there other things that you can do for integrators, such as SPC alarming after the fact or does Zebra look at the integrator and say, We want to help you solve your customers' problems today but we also want to see if we can maybe open up some revenue streams down the road while making the system even more bulletproof? I mean, just making it easier to maintain and support.

Fabio Perelli:
There's a lot of talk about monitoring your system in the field, making sure that it's performing as it's supposed to and then start figuring out when things are not as sharp as they should be, as you can say, and then provide some information, then let the integrator go in the field, service, and do preventive maintenance as opposed to, Oh, the thing broke and now the line is down for a week because I didn't find the replacement. So if you're able to see ahead when your system does need some type of maintenance work, well then, and not just your inspection system, obviously, but mostly the machine that's working the rest of the application. Because the vision most likely is a smaller portion of a larger system.

Winn Hardin:
I was always so impressed the first time I went to FANUC's campus and saw their ZDT control center. Just that ability to predict failure and to react so that you don't have unplanned downtime, because that's the most expensive aspect.

Fabio Perelli:
And again, with deep learning you're able to forecast a lot better some of these modes of failures, when they could happen. Or you can look at trends and understand, okay, this is not going in the right direction, something is going to fail eventually. So then preventive maintenance is the key there.

Winn Hardin:
Yes, I don't think we talked enough about deep learning's ability with SPC, to really make SPC come alive and provide that intelligence, so that you're not having to sift through reams and reams of data to figure out, and waste your time. You don't have to go to it if there's not a problem.

Fabio Perelli:
So yeah, because in-field service is an expensive . . .

Winn Hardin:
proposition for everybody, especially when you have to fly them out.

Jimmy Carroll:
One thing I wanted to ask about that I was interested in personally was on the hardware side, in 3D. So obviously now you have 3D capabilities with the LTs. Am I saying that right?

Fabio Perelli:
LTs, yes.

Jimmy Carroll:
So obviously you guys have 3D imaging capabilities now. If we use the warehouse as an example or in the packaging, there are clear applications for 3D volumetric measurements of food and packaging and correct fill line measurements or whatever else. But are you seeing particular interest from different industries for your 3D sensor in particular? Or is it largely in that packaging, warehouse . . .

Fabio Perelli:
We're doing a lot work, actually, with some of the battery manufacturers for electric vehicles. When you assemble all these layers of anodes, cathodes, and other things, precision is really the key. And if you remember those Samsung fires in the phones. That was mostly caused by misalignments of things and things were not assembled properly. So there's a lot of inspection actually done in 3D with the LTs for a lot of these car batteries to make sure that there's nothing that's out of tolerance, and then you try to fit it in . . .

Winn Hardin:
Short, bang, fire.

Fabio Perelli:
So it doesn't take much apparently to get issues. So we are working heavily in that industry these days with the LTs, and again 3D is by far one of the largest places where we're developing new functionality, not just on the software side but also on the product side. So let me say this properly. The Aurora Imaging Library, which is our old Matrox Imaging Library, we're adding a lot of 3D functionality in it, because obviously, yes, you can get 3D data Well you have to do something useful with the data to provide a whole solution. So that's happening. And we're also working on other devices to complement the LTs. So the LT is a profile sensor. So we have a laser line you know. So you need motion to create a 3D profile and a point cloud. We're looking now at a snapshot sensor so that you can get a 3D image without having that relative motion, because you don't have that all the time.

Winn Hardin:
In all the applications all the time.

Fabio Perelli:
So we're looking at that and other devices, like time of flight. And we see that this is definitely by far the highest growth market. 3D is really taking off. We can solve applications that are just impossible with the 2D vision, no matter what type of lighting. You can't do it. Whereas with 3D it becomes fairly straightforward.

Winn Hardin:
You're doing most of it in software.

Fabio Perelli:
And then you have the software back end that takes in all this data and does the analysis in three dimensions. A lot of our software does a lot of the inspection portion once you have the 3D data. Again, automotive is one of the industries where we see a lot of demand. People need to inspect these connectors for all the cable harnesses. It looks simple between me and you to look at something and say, oh, that looks a little bit different. For 2D vision, it's very difficult. You add that third dimension and it becomes obvious. And this is where we have solutioned a lot of issues and helped a lot of manufacturers with this kind of technology. Definitely a growth market.

Winn Hardin:
Large-area three-dimensional inspection at speed is an impressive feat. It's just like when we made the transition from monochrome to color and then went from color to 3D. Price was always a consideration that slowed growth a little bit, but it sounds like you guys are overcoming some of that, which is bringing it to a larger audience.

Fabio Perelli:
And that's why we're working on a range of 3D sensors, because depending on the technology that you use, there are different costs involved. And not everybody needs the top of everything. You can get away with lower-cost solutions because your application doesn't need that much precision or tolerance or whatever, or that kind of speed. So that's why we're trying to diversify beyond just the simple laser profiler to give more options to our customers, our integrators.

Winn Hardin:
When you cut the laser out, it also takes away some of the safety issues, some of the environmental issues, concerns that people might have, with humans working nearby. So that's impressive stuff.

Jimmy Carroll:
One trend I've noticed lately is the rise or the increased availability of application-specific systems, these purpose-built systems that incorporate a robot and software, oftentimes AI based, along with 3D. Often for bin picking or palletizing, depalletizing — specific applications. I'm just curious, are you seeing LTs being used in this way?

Fabio Perelli:
LTs possibly not that much. Especially for the bin picking, the snapshot sensor technology is probably a better technology. On the other hand, for picking things off a conveyor, LT is perfect because you already have the relative motion. So it's perfect. People use that with our product. But again, we're trying to diversify beyond the laser profiler so that I can offer more and more options.

Winn Hardin:
A question that was in my head, Fabio: Is Zebra interested in moving into an embedded vision space? But also we talked about warehousing, we were talking about 3D and warehousing, and we were talking about the battery manufacture and assembly. I was assuming it was going to be in a palletizer or a cartoner or something. that would be relevant. But is there any interest in moving into, like, the autonomous forklift, the AMR markets, other folks?

Fabio Perelli:
I think it's probably a natural progression right in there. I mean the next thing people were looking for is just computer volume on box. I need to understand how big it is so that I can see is it going to fit in my truck or I have to go in the next one. That kind of stuff. You have a scan tunnel set up already to read all kinds of other information, and why don't you make the measurement at the same time? It's all set up there. Just add an extra instrument. We have the software to do all of that and tie it all together. So this is the, as I was saying, the excitement that I had when Zebra and Matrox joined forces because now we have two worlds that come together into this super solution. So I think it's quite, quite exciting in that respect.

Winn Hardin:
Sure. I love the modern warehouses. It's one of my favorite installations to go visit. I was up at Dick's Sporting Goods in Conklin, New York, a couple of years ago, and at AutoStore Cube as well. It's just full sorting all the way across three different buildings. To me, it's amazing. An Amazon warehouse is the big example, although a lot of us don't get to walk in there.

Jimmy Carroll:
And it's cool too because you get to see all the different types of machine vision technologies. And there's so many different applications within the warehouse. So as all these technologies progress and advance, you get to see them in the warehouse, like working together. And it's very cool. And on that note, Fabio, in the general automation space, not just in the warehouse but in the general industrial automation manufacturing space, are there any trends or predictions that you see as likely in the next couple of years?

Fabio Perelli:
I mean, the robotic is going to continue.

Winn Hardin:
To continue to grow and expand.

Fabio Perelli:
Difficulty of getting manpower in general. We have no choice. We need to get robots to come in. The AMR market, I think, is going to evolve even further. People don't want to carry things or go run around in a warehouse. Robots are good at that. They can do that without even light. Perfect. So why don't we push more of that? Having a robot running around your Walmart aisle as you're shopping may be a bit intimidating. For the warehousing, the back end, why not?

Winn Hardin:
I completely agree, and we'd already seen a philosophical change 10, 15 years ago, when it was: put the assemblers in one location and bring the parts to them, right, as opposed to having them go get the parts. More efficient operations and everything. If you can combine that with AMRs to then keep their bins full and their part availability and right next to them, we're just going to continue to see more productivity, which will help to drive that interest we're talking about. Buy more bots and more cameras.

Fabio Perelli:
Correct. And obviously Zebra had that long-term view because you know the acquisition of Fetch was definitely . . . one of the places where we want to be is in factory automation at the AMR level. And bring all of the vision stuff on the other side as well to add on. Because not only do you want to be able to read a product; you want to look at a product, see if it's the right piece of equipment that's tagged with such and such a barcode, and make sure also that the box is not damaged. There's a lot of applications where machine vision can sit on top of all these other technologies all at once. So you're providing a global solution as opposed to having individual systems that each one does its own little bit. And then you have to all tie it together. We can tie it together, provide a solution.

Winn Hardin:
Does Aurora already have that connectivity with the Fetch product line?

Fabio Perelli:
A little bit, but obviously there's more work to be put in there. One of the things also that I should mention is on the industrial scanning stuff, we are doing a lot of what we call ready solutions. So we're bundling the tools and the software, hardware package that an integrator might need to do dock door inspection, when goods, pallets come in and out. We're going to give you basically a recipe, chef, on how to solution this application so that you don't need to run out and figure out all kinds of things. We do it for you. You just set it up physically and then you maintain it. So then you get your service contract and all the things we've discussed before.

Jimmy Carroll:
And you know which ingredient is going to go together to make really good soup.

Fabio Perelli:
And we know exactly how to make the right cake so that you don't make a mixup. So we have a couple of these ready solutions on the industrial scanning, and we're going to keep on adding more and more to the machine vision portfolio as well, because there's some machine vision applications that are not identical, but they're repeatable, with some little tweaks here and there, but you have a basis to start your solution.

Winn Hardin:
It's primarily focused on the sorting and the warehouse, the packaging area you were talking about before?

Fabio Perelli:
Potentially just even inspection in general. There's a lot of things that get inspected, irrelevant if it's like a pencil or a or marker. There's similarities there. So you can provide a solution that can do that.

Winn Hardin:
Absolutely. Absolutely.

Jimmy Carroll:
Fabio, anything else that we haven't asked that you want to make mention of today?

Fabio Perelli:
Well, I would keep on looking at us because we have a lot of things in our road map, a lot of new things are going to come out in the first half of this year. And talking in terms of these 3D solutions, we are finishing up all these new products that are going to come out. Software we're going to keep on evolving. This has been ongoing for many, many years. But we're going to keep on enhancing. And we are benefiting now from having the ability to pick and choose the best from our libraries, which we have from the Adaptive Vision and from the Matrox side. We can pick and choose to then put into the Design Assistant or in the Focus software. So we have a lot of flexibility.

Winn Hardin:
We're starting to see those combined solutions where you're pulling in from the different acquisitions. I mean that really speaks more than just to this interview about the level of integration that you've been able to bring together.

Fabio Perelli:
So all the deep learning, OCR. Well, that comes from the Adaptive Vision. We have coming out very, very soon a generalized anomaly detection, which is something that I think is really going to finally jump-start the DL adoption. Because I travel many times to see customers with standard classification tools. Yes, they are deep learning based, but you always need a large number of images of good parts and bad parts.

Winn Hardin:
Very hard to get.

Fabio Perelli:
So everybody tells you I have lots of good parts. That's my business. I don't make bad parts. So for me to give you even 100th of the good images . . .

Winn Hardin:
It's not realistic. Especially when you need it for every single defect class.

Fabio Perelli:
Correct. You need to do the opposite and all the detection. This is the good part, what it should look like. As soon as it deviates, well there's something wrong. So now you're basically making the network. You're training it with the good images.

Winn Hardin:
It'll be interesting to see how you achieve just the right threshold there, because that's going to be set with every application. But then you need to guide that integrator and that designer.

Fabio Perelli:
Correct. So I think the other place where it's going to be a lot of development is on the training tools.

Winn Hardin:
I agree.

Fabio Perelli:
Because you cannot be a PhD person, that knows how to tweak the kernels and all that.

Winn Hardin:
Identify the outliers that are destroying your model.

Fabio Perelli:
Correct. So you need to find tools that are very simple to use and not just to do it once, how to then feed it back and retrain it with outliers, issues that came up that you were not forecasting. So this is the place where we also need to do a lot of work.

Winn Hardin:
Cool. I expect the pace of change will probably be accelerating too. I mean as you guys really start to bring all these companies' technologies together.

Fabio Perelli:
And we have that advantage, right. Because we can share technology within the Zebra organization, which is quite, quite powerful. We have advantages there.

Jimmy Carroll:
Well Fabio, I certainly look forward to seeing what Zebra has to offer in the coming months and this year. And I want to thank you for your time. It's been a pleasure to talk to you. For any questions or comments or if anyone wants to reach out, you can do so at Manufacturing dash Matters.com and you can learn more about Zebra at Zebra.com. And thanks everyone for joining.

Jimmy Carroll:
Well, thank you very much for your time. Thank you.

Sonix is the world’s most advanced automated transcription, translation, and subtitling platform. Fast, accurate, and affordable.

Automatically convert your m4a files to text (txt file), Microsoft Word (docx file), and SubRip Subtitle (srt file) in minutes.

Sonix has many features that you’d love including world-class support, powerful integrations and APIs, upload many different filetypes, enterprise-grade admin tools, and easily transcribe your Zoom meetings. Try Sonix for free today.

Jimmy Carroll: [00:00:06] Hi, everybody. My name is Jimmy Carroll. I’m the vice president of operations at Tech B2B Marketing. I’m here for the Manufacturing Matters podcast. I have the pleasure of being joined by my colleague Winn Hardin and Fabio Pirelli from Zebra. Fabio, thank you so much for taking the time. It’s always a pleasure to see you. For those who don’t know, if you could give us a quick background on Zebra and what you do there, I think that would be helpful.

Fabio Perelli: [00:00:27] Sure. Okay, so I’m a product manager. I work in the product team. I have a team of guys that help me out, and basically we manage all of the hardware development on the machine vision side of things. And you also probably know my counterpart Pierantonio. He also looks at all the software, because today if you really look at the products, you have a combination always of software and hardware. A piece of hardware on its own, it doesn’t really do that much. So you need to have this combination. So Pierantonio and I co-jointly develop and follow up development of all the machine vision products there. 

Winn Hardin: [00:01:06] Now Zebra does a lot of other things too in addition to machine vision.

Fabio Perelli: [00:01:09] Correct. So we are part of the Machine Vision and Fixed Industrial Scanner Group. Zebra has a division that does printers. Then you know the end health scanners, RFID, many, many other products, but the group I’m in is machine vision–centric.

Winn Hardin: [00:01:30] Well that’s good. You’ve done that cross-pollination between the different product lines. It makes you stronger as a company in terms of your offerings for sure.

Fabio Perelli: [00:01:35] Well, it opens doors. You sell RFID and well, maybe you want to inspect your products.

Jimmy Carroll: [00:01:41] I want to ask about that. And I’m going to ask about some specific technologies within the Machine Vision Group. But before I do, I have to ask, how has the transition from Matrox to Zebra gone and what’s most exciting about it to you?

Fabio Perelli: [00:01:55] The transition is I would say pretty much completed. There’s always a little hiccup here and there, but overall I think everybody did a great job. It was a pretty big undertaking that took quite a bit of time, but I think it worked out. We’re pretty much a cohesive team at the moment, so it’s very, very good. One of the things that got us very excited when we got the news that we were joining Zebra was the fact that we had very complementary product lines. And I’m talking mostly, again, in the machine vision business units. They had a smart camera, yes. And they had industrial scanners, yes. We had a smart camera too, but it was at a somewhat different level. And the software offering was very different. On the other hand, today we have lots of software to offer, which is a great advantage because if one package doesn’t do something, we have something else to offer as a complementary product, which is nice. Library tools — we’ve got mountains and mountains of functionality there, so I think we’re in a pretty good position.

Winn Hardin: [00:03:11] Back in the day, machine vision systems would almost be standalone nodes, but that’s not the way it is now with IoT, Industry 4.0. And that’s not just buzzwords anymore. I mean, we’re really seeing that communication across, and not just in the plant now, but as the cloud, and I would love to hear what you guys think about the cloud. And there’s a lot of discussion about whether this is going to be the year of the cloud. The PLCs, the other programs are going to start having that embedded in, it being more of a requirement. I’ve always been a bit of a skeptic, because manufacturers want to protect their data at all costs. So it’s a hard, difficult hurdle to overcome a little bit from acceptance.

Fabio Perelli: [00:03:51] Yeah, the cloud has been a lot of talk for, as you said, many years. A lot of our customers are still somewhat nervous about it because their data, which is their secret, is transiting who knows where. On the other hand, what we have seen for the deep learning or AI stuff, you bring the cloud on-site. So you actually set up a computer farm right on your facility. So you get all the benefit of the big computing, but then you keep everything local and some other ones you’re able to encrypt data. OPC UA is a good example, where everything is encrypted, A to Z, for the protocol, hardware, everything is encrypted. You should not be as concerned in my opinion. So yeah, the facilities could be used without too much worrying about losing or competitors taking bits and pieces of your secret sauce there. 

Jimmy Carroll: [00:04:54] AI is something I wanted to talk about. So Zebra, there’s a lot to talk about in the Machine Vision Group, from barcode scanning and machine vision cameras and smart cameras. And we could talk about this for hours, but one of the topics I want to ask about is deep learning. So it’s a buzzword I understand. But in the last couple of years, it’s really seemed to become a little bit more of a mature technology. And as people start to realize that it’s a tool for the machine vision toolbox, it’s not a magic tool that’s going to solve everything, but there are very useful, practical applications for it. And one of those is OCR. So your tool is designed around OCR, designed for OCR. How can this make the vision and the vision engineer or the systems integrator’s job easier?

Fabio Perelli: [00:05:39] Yes. A very good question. I mean, overall, the deep learning had been more of a buzzword, a lot of science projects, PhD thesis. I think now we’re at the moment where we can actually physically use it in practical applications to solve real problems. What you’re discussing here is what we call the deep learning–based OCR tool. It’s pretty powerful. I mean, I’ve been in the industry for 27-odd years — unfortunately you can you tell my age. On the other hand, I did see a lot of these issues — reading, texts, and all that stuff. This deep learning–based OCR tool, it really freaked me out because it works. And it works out of the box. There’s no complex training, which with a lot of the previous versions of OCR tools, you had to get the alphabet out explicitly, say this is an A, a B, a C, and it takes forever. And then you get a slight variation and then you have to start over again. This thing, you just get an image with text in it, and it reads it, irrelevant of the size, the shape. It was pretty impressive.

Winn Hardin: [00:06:58] And for me, deploying deep learning, one of the biggest hurdles has been getting the integrators. Because the integrators got burned on a lot of the early — I mean, forget about the neural network stage, but when AI/deep learning was first coming out, a lot of people spent a lot of time trying to solve a lot of applications with it and were not always as successful as they wanted, to see that ROI. So I would think that if you’re able to  use it in a finite space, like in an OCR application, and it can continue to learn and evolve in a continuous learning environment, that seems like maybe it’ll start to create that confidence as well as the familiarity.

Fabio Perelli: [00:07:36] Again because training a network is not that straightforward. So we give you something that’s pre-trained, with potentially millions of images that we used to get to that point. So obviously when you show something it will probably work right out of the box. So then therefore the integrator has very little time to waste trying to get everything to work. He’s almost guaranteed that it’s going to work out of the box without any crazy lighting setup.

Winn Hardin: [00:08:07] And then three to six months putting a temporary system on the line so you can acquire the datasets.

Fabio Perelli: [00:08:12] Exactly. So you bypass and jump over all that problematic stage, which is very time-consuming. Costs money. To have something that you can solution and show the customer: This is how it works and it works quite well. At Automate, I don’t know if you remember, we had a demo with a rotating table that had four different stations of four things that were totally different. One was like metallic dot pin text. It reads it, with the same lighting as when you’re trying to read something on a curved bottle. And there was something and there was like inverse contrastive . . . 

Winn Hardin: [00:08:53] We’re talking nonreflective surfaces in that particular case, which is really astounding.

Fabio Perelli: [00:08:54] Embossed, I mean, it really, really, really solutioned the problem without having to create special lighting for each one of these scenarios. Again, I was very impressed with that because typically you need to do a lot more work on that.

Winn Hardin: [00:09:09] You’re building that where it combines photometric with standard two dimensional, and I think three dimensional is something else you wanted to talk about today too. I mean, does it come with a standard optical head? 

Fabio Perelli: [00:09:21] No, typically, depending on what product you’re going to choose. If you choose, for example, the RS-GTX, you’re free to pick your own lenses and your own lighting so you can set up whatever distance you want it to work, whatever lighting conditions and all that. We also have a line of what we call the VS cameras, which came from the Zebra side. Well, some of them have the complete illumination built in. You have liquid lenses for focusing. So a more complete package. It works, maybe not in every condition, but yes, it’s easier to set up for an integrator, for example. And the algorithm can run in either one of these platforms. So all of a sudden you have the flexibility of choosing very different and variable forms with software that’s kind of common, so you know that’s going to work, irrelevant if you pick A or B, because it all comes from the same . . . 

Winn Hardin: [00:10:19] And that’s the true power . . . I mean, we’ve rarely seen enough of that in machine vision, where a software can be truly hardware agnostic. I mean, it’s usually going to be very much a paired system between the software, because machine vision is difficult to be successful at. But it’s really cool to see software evolving, whether it’s an IDE with an AI OCR tool built into it, to basically say: You can run on any of our platforms. But frankly, if you’re a Ford or GM and you’re married to certain other hardware platforms, our software can still be the powerhouse that makes you successful.

Fabio Perelli: [00:10:57] A lot of flexibility. 

Jimmy Carroll: [00:10:59] I think it speaks to the whole idea of ease of use. That’s one of the terms that keeps emerging. And maybe it’s an obvious one. Everything needs to become easier to use. But do you see that as a main growth driver in the increased adoption of automation? for example, obviously this is a product along with many of your others that can be deployed in warehouses. So warehouse automation market intelligence company Interact Analysis predicts modest growth for warehouse automation in 2024 but double-digit growth in the years to follow. Do you see ease of use as one of the main growth drivers there? And what else do you see?

Fabio Perelli: [00:11:33] I think ease of use is definitely one of the key things that we are also working on. We’re trying to get that learning curve of learning a software or understanding how things work as simple as possible. Because that’s time wasted, because people want to solution things. They don’t want to learn. People know the vision. They understand vision. They don’t need to understand how our package works. They just want to use it in the best possible way to get a solution up and running. So this is definitely one of the focuses that we have. And, pardon the pun, but the Aurora Focus, which is what Zebra has developed for the smart camera line and for the fixed industrial scanners, it’s very simple to use. You can get up and running very quickly, and we’re going to evolve this more and more. We’re going to continue developing that. And we also have the Aurora Design Assistant, which comes from the Matrox side of the business, where if you need more complex, more evolved type of functionality, you have that as well. 

Winn Hardin: [00:12:42] That’s really what I wanted to ask, because you can have super-powerful algorithms, quote unquote idiot-proof algorithms. But it takes more. When you talk about ease of use, I was going to push you on that Fabio a little bit, just because everybody says our software is easy to use, but they don’t really talk about illustrative examples, all the different little parts. So Design Assistant is a critical element, as well as the powerful algorithms behind it. Are there other aspects in terms of, and I know it’s like talking about machine vision and then talking about training, you rarely do that. But in terms of knowledge bases, in terms of in-line help systems, which the Design Assistant provides part of that. But are there other things that Zebra does to truly make that easy-to-use software? Is it primarily the algorithms and the Design Assistant or is it more of a whole, also, with what you’ve got on your website, you have educational programs? 

Fabio Perelli: [00:13:38] We do have a portion of the website that has training, online training. So you can self-train. We offer physical training courses. We offer online courses. So there’s a whole slew of ways to get familiarized with how to use the software. We’re looking at how to leverage some of the ChatGTP type of artificial intelligence to cater to . . . 

Winn Hardin: [00:14:09] We might as well eat our own cooking.

Fabio Perelli: [00:14:11] So something that we’re looking into is an easy way to — you ask a question in plain English and you get . . . 

Winn Hardin: [00:14:21] the answer that you truly need.

Fabio Perelli: [00:14:22] Correct.

Winn Hardin: [00:14:23] And that’s one of the real benefits of working with one of the larger companies, especially machine vision, which is traditionally been a lot of little companies with a couple of monsters in there. You have the resources to be able to offer all these systems. So the training isn’t an afterthought, so that aftermarket support is there without always having to push back to the integrator. Integrators don’t always prioritize those maintenance contracts long term, which is always something over the years that has surprised me. But I guess we can just make more revenue. But I wonder, as far as when you’re designing and solving an application, that’s obviously super-critical, but are there other things that you can do for integrators, such as SPC alarming after the fact or does Zebra look at the integrator and say, We want to help you solve your customers’ problems today but we also want to see if we can maybe open up some revenue streams down the road while making the system even more bulletproof? I mean, just making it easier to maintain and support.

Fabio Perelli: [00:15:27] There’s a lot of talk about monitoring your system in the field, making sure that it’s performing as it’s supposed to and then start figuring out when things are not as sharp as they should be, as you can say, and then provide some information, then let the integrator go in the field, service, and do preventive maintenance as opposed to, Oh, the thing broke and now the line is down for a week because I didn’t find the replacement. So if you’re able to see ahead when your system does need some type of maintenance work, well then, and not just your inspection system, obviously, but mostly the machine that’s working the rest of the application. Because the vision most likely is a smaller portion of a larger system. 

Winn Hardin: [00:16:16] I was always so impressed the first time I went to FANUC’s campus and saw their ZDT control center. Just that ability to predict failure and to react so that you don’t have unplanned downtime, because that’s the most expensive aspect.

Fabio Perelli: [00:16:30] And again, with deep learning you’re able to forecast a lot better some of these modes of failures, when they could happen. Or you can look at trends and understand, okay, this is not going in the right direction, something is going to fail eventually. So then preventive maintenance is the key there. 

Winn Hardin: [00:16:48] Yes, I don’t think we talked enough about deep learning’s ability with SPC, to really make SPC come alive and provide that intelligence, so that you’re not having to sift through reams and reams of data to figure out, and waste your time. You don’t have to go to it if there’s not a problem. 

Fabio Perelli: [00:17:03] So yeah, because in-field service is an expensive . . .

Winn Hardin: [00:17:06] proposition for everybody, especially when you have to fly them out. 

Jimmy Carroll: [00:17:11] One thing I wanted to ask about that I was interested in personally was on the hardware side, in 3D. So obviously now you have 3D capabilities with the LTs. Am I saying that right? 

Fabio Perelli: [00:17:22] LTs, yes. 

Jimmy Carroll: [00:17:24] So obviously you guys have 3D imaging capabilities now. If we use the warehouse as an example or in the packaging, there are clear applications for 3D volumetric measurements of food and packaging and correct fill line measurements or whatever else. But are you seeing particular interest from different industries for your 3D sensor in particular? Or is it largely in that packaging, warehouse . . . 

Fabio Perelli: [00:17:52] We’re doing a lot work, actually, with some of the battery manufacturers for electric vehicles. When you assemble all these layers of anodes, cathodes, and other things, precision is really the key. And if you remember those Samsung fires in the phones. That was mostly caused by misalignments of things and things were not assembled properly. So there’s a lot of inspection actually done in 3D with the LTs for a lot of these car batteries to make sure that there’s nothing that’s out of tolerance, and then you  try to fit it in . . .

Winn Hardin: [00:18:33] Short, bang, fire. 

Fabio Perelli: [00:18:37] So it doesn’t take much apparently to get issues. So we are working heavily in that industry these days with the LTs, and again 3D is by far one of the largest places where we’re developing new functionality, not just on the software side but also on the product side. So let me say this properly. The Aurora Imaging Library, which is our old Matrox Imaging Library, we’re adding a lot of 3D functionality in it, because obviously, yes, you can get 3D data Well you have to do something useful with the data to provide a whole solution. So that’s happening. And we’re also working on other devices to complement the LTs. So the LT is a profile sensor. So we have a laser line you know. So you need motion to create a 3D profile and a point cloud. We’re looking now at a snapshot sensor so that you can get a 3D image without having that relative motion, because you don’t have that all the time.

Winn Hardin: [00:19:42] In all the applications all the time.

Fabio Perelli: [00:19:44] So we’re looking at that and other devices, like time of flight. And we see that this is definitely by far the highest growth market. 3D is really taking off. We can solve applications that are just impossible with the 2D vision, no matter what type of lighting. You can’t do it. Whereas with 3D it becomes fairly straightforward.

Winn Hardin: [00:20:11] You’re doing most of it in software.

Fabio Perelli: [00:20:13] And then you have the software back end that takes in all this data and does the analysis in three dimensions. A lot of our software does a lot of the inspection portion once you have the 3D data. Again, automotive is one of the industries where we see a lot of demand. People need to inspect these connectors for all the cable harnesses. It looks simple between me and you to look at something and say, oh, that looks a little bit different. For 2D vision, it’s very difficult. You add that third dimension and it becomes obvious. And this is where we have solutioned a lot of issues and helped a lot of manufacturers with this kind of technology. Definitely a growth market.

Winn Hardin: [00:21:03] Large-area three-dimensional inspection at speed is an impressive feat. It’s just like when we made the transition from monochrome to color and then went from color to 3D. Price was always a consideration that slowed growth a little bit, but it sounds like you guys are overcoming some of that, which is bringing it to a larger audience.

Fabio Perelli: [00:21:24] And that’s why we’re working on a range of 3D sensors, because depending on the technology that you use, there are different costs involved. And not everybody needs the top of everything. You can get away with lower-cost solutions because your application doesn’t need that much precision or tolerance or whatever, or that kind of speed. So that’s why we’re trying to diversify beyond just the simple laser profiler to give more options to our customers, our integrators. 

Winn Hardin: [00:21:57] When you cut the laser out, it also takes away some of the safety issues, some of the environmental issues, concerns that people might have, with humans working nearby. So that’s impressive stuff. 

Jimmy Carroll: [00:22:08] One trend I’ve noticed lately is the rise or the increased availability of application-specific systems, these purpose-built systems that incorporate a robot and software, oftentimes AI based, along with 3D. Often for bin picking or palletizing, depalletizing — specific applications. I’m just curious, are you seeing LTs being used in this way?

Fabio Perelli: [00:22:36] LTs possibly not that much. Especially for the bin picking, the snapshot sensor technology is probably a better technology. On the other hand, for picking things off a conveyor, LT is perfect because you already have the relative motion. So it’s perfect. People use that with our product. But again, we’re trying to diversify beyond the laser profiler so that I can offer more and more options.

Winn Hardin: [00:23:06] A question that was in my head, Fabio: Is Zebra interested in moving into an embedded vision space? But also we talked about warehousing, we were talking about 3D and warehousing, and we were talking about the battery manufacture and assembly. I was assuming it was going to be in a palletizer or a cartoner or something. that would be relevant. But is there any interest in moving into, like, the autonomous forklift, the AMR markets, other folks?

Fabio Perelli: [00:23:35] I think it’s probably a natural progression right in there. I mean the next thing people were looking for is just computer volume on box. I need to understand how big it is so that I can see is it going to fit in my truck or I have to go in the next one. That kind of stuff. You have a scan tunnel set up already to read all kinds of other information, and why don’t you make the measurement at the same time? It’s all set up there. Just add an extra instrument. We have the software to do all of that and tie it all together. So this is the, as I was saying, the excitement that I had when Zebra and Matrox joined forces because now we have two worlds that come together into this super solution. So I think it’s quite, quite exciting in that respect.

Winn Hardin: [00:24:20] Sure. I love the modern warehouses. It’s one of my favorite installations to go visit. I was up at Dick’s Sporting Goods in Conklin, New York, a couple of years ago, and at AutoStore Cube as well. It’s just full sorting all the way across three different buildings. To me, it’s amazing. An Amazon warehouse is the big example, although a lot of us don’t get to walk in there.

Jimmy Carroll: [00:24:41] And it’s cool too because you get to see all the different types of machine vision technologies. And there’s so many different applications within the warehouse. So as all these technologies progress and advance, you get to see them in the warehouse, like working together. And it’s very cool. And on that note, Fabio, in the general automation space, not just in the warehouse but in the general industrial automation manufacturing space, are there any trends or predictions that you see as likely in the next couple of years?

Fabio Perelli: [00:25:09] I mean, the robotic is going to continue.

Winn Hardin: [00:25:11] To continue to grow and expand.

Fabio Perelli: [00:25:13] Difficulty of getting manpower in general. We have no choice. We need to get robots to come in. The AMR market, I think, is going to evolve even further. People don’t want to carry things or go run around in a warehouse. Robots are good at that. They can do that without even light. Perfect. So why don’t we push more of that? Having a robot running around your Walmart aisle as you’re shopping may be a bit intimidating. For the warehousing, the back end, why not?

Winn Hardin: [00:25:48] I completely agree, and we’d already seen a philosophical change 10, 15 years ago, when it was: put the assemblers in one location and bring the parts to them, right, as opposed to having them go get the parts. More efficient operations and everything. If you can combine that with AMRs to then keep their bins full and their part availability and right next to them, we’re just going to continue to see more productivity, which will help to drive that interest we’re talking about. Buy more bots and more cameras.

Fabio Perelli: [00:26:17] Correct. And obviously Zebra had that long-term view because you know the acquisition of Fetch was definitely . . . one of the places where we want to be is in factory automation at the AMR level. And bring all of the vision stuff on the other side as well to add on. Because not only do you want to be able to read a product; you want to look at a product, see if it’s the right piece of equipment that’s tagged with such and such a barcode, and make sure also that the box is not damaged. There’s a lot of applications where machine vision can sit on top of all these other technologies all at once. So you’re providing a global solution as opposed to having individual systems that each one does its own little bit. And then you have to all tie it together. We can tie it together, provide a solution.

Winn Hardin: [00:27:05] Does Aurora already have that connectivity with the Fetch product line?

Fabio Perelli: [00:27:09] A little bit, but obviously there’s more work to be put in there. One of the things also that I should mention is on the industrial scanning stuff, we are doing a lot of what we call ready solutions. So we’re bundling the tools and the software, hardware package that an integrator might need to do dock door inspection, when goods, pallets come in and out. We’re going to give you basically a recipe, chef, on how to solution this application so that you don’t need to run out and figure out all kinds of things. We do it for you. You just set it up physically and then you maintain it. So then you get your service contract and all the things we’ve discussed before. 

Jimmy Carroll: [00:28:01] And you know which ingredient is going to go together to make really good soup.

Fabio Perelli: [00:28:03] And we know exactly how to make the right cake so that you don’t make a mixup. So we have a couple of these ready solutions on the industrial scanning, and we’re going to keep on adding more and more to the machine vision portfolio as well, because there’s some machine vision applications that are not identical, but they’re repeatable, with some little tweaks here and there, but you have a basis to start your solution.

Winn Hardin: [00:28:32] It’s primarily focused on the sorting and the warehouse, the packaging area you were talking about before?

Fabio Perelli: [00:28:36] Potentially just even inspection in general. There’s a lot of things that get inspected, irrelevant if it’s like a pencil or a or marker. There’s similarities there. So you can provide a solution that can do that.

Winn Hardin: [00:28:51] Absolutely. Absolutely.

Jimmy Carroll: [00:28:54] Fabio, anything else that we haven’t asked that you want to make mention of today?

Fabio Perelli: [00:28:59] Well, I would keep on looking at us because we have a lot of things in our road map, a lot of new things are going to come out in the first half of this year. And talking in terms of these 3D solutions, we are finishing up all these new products that are going to come out. Software we’re going to keep on evolving. This has been ongoing for many, many years. But we’re going to keep on enhancing. And we are benefiting now from having the ability to pick and choose the best from our libraries, which we have from the Adaptive Vision and from the Matrox side. We can pick and choose to then put into the Design Assistant or in the Focus software. So we have a lot of flexibility.

Winn Hardin: [00:29:51] We’re starting to see those combined solutions where you’re pulling in from the different acquisitions. I mean that really speaks more than just to this interview about the level of integration that you’ve been able to bring together.

Fabio Perelli: [00:30:01] So all the deep learning, OCR. Well, that comes from the Adaptive Vision. We have coming out very, very soon a generalized anomaly detection, which is something that I think is really going to finally jump-start the DL adoption. Because I travel many times to see customers with standard classification tools. Yes, they are deep learning based, but you always need a large number of images of good parts and bad parts. 

Winn Hardin: [00:30:40] Very hard to get.

Fabio Perelli: [00:30:41] So everybody tells you I have lots of good parts. That’s my business. I don’t make bad parts. So for me to give you even 100th of the good images . . .  

Winn Hardin: [00:30:54] It’s not realistic. Especially when you need it for every single defect class. 

Fabio Perelli: [00:30:58] Correct. You need to do the opposite and all the detection. This is the good part, what it should look like. As soon as it deviates, well there’s something wrong. So now you’re basically making the network. You’re training it with the good images.

Winn Hardin: [00:31:11] It’ll be interesting to see how you achieve just the right threshold there, because that’s going to be set with every application. But then you need to guide that integrator and that designer.

Fabio Perelli: [00:31:20] Correct. So I think the other place where it’s going to be a lot of development is on the training tools.

Winn Hardin: [00:31:27] I agree.

Fabio Perelli: [00:31:28] Because you cannot be a PhD person, that knows how to tweak the kernels and all that. 

Winn Hardin: [00:31:35] Identify the outliers that are destroying your model. 

Fabio Perelli: [00:31:38] Correct. So you need to find tools that are very simple to use and not just to do it once, how to then feed it back and retrain it with outliers, issues that came up that you were not forecasting. So this is the place where we also need to do a lot of work.

Winn Hardin: [00:31:57] Cool. I expect the pace of change will probably be accelerating too. I mean as you guys really start to bring all these companies’ technologies together.

Fabio Perelli: [00:32:03] And we have that advantage, right. Because we can share technology within the Zebra organization, which is quite, quite powerful. We have advantages there. 

Jimmy Carroll: [00:32:16] Well Fabio, I certainly look forward to seeing what Zebra has to offer in the coming months and this year. And I want to thank you for your time. It’s been a pleasure to talk to you. For any questions or comments or if anyone wants to reach out, you can do so at Manufacturing dash Matters.com and you can learn more about Zebra at Zebra.com. And thanks everyone for joining.

Jimmy Carroll: [00:32:36] Well, thank you very much for your time. Thank you.