Fever Detecting Cameras – Hot Tech or Hot Air?

 

 

 

 

Transcript of Podcast.

Daniel: This is Daniel Lewkovitz from Calamity. Since coronavirus saw businesses shut down, there has been an urgent desire for society to reopen as quickly as possible. As we adjust to what’s been called the “new normal,” a key aim for government and the private sector is to stop the spread of coronavirus and infection in general. Virtually overnight, the security industry has started promoting thermal fever detecting CCTV cameras in a big way, as a possible solution of preventing infected individuals coming on to protected properties. This technology is now being marketed as a solution to event organisers at large venues as well as hospitals, nursing homes, schools, and businesses, big and small. But does it actually work? Well, I’m not convinced. Best case, even if the technology did work and it was foolproof, one area that has not been given much consideration is: what do organisations do next? 

Do they have security protocols and resources in place to properly manage such a detection by system? What do they do? Eject people? How they are going to manage the privacy concerns arising from all of this? And what would happen if people sought to deliberately defeat such as a system?

For example, a young person who was determined to see a once-in-a-lifetime concert. So they splash some cold water on their forehead to lower the temperature before entering a venue that they believe was going to be temperature-screening individuals. It’s a modern take on concealing drugs before you enter a dance party. Or what about when people’s temperature might affect their ability to attend work and earn wages? The New York Times has already reported of an Iowa meat processing plant worker who died of coronavirus. Now, they had been taking Tylenol to reduce their temperature in a deliberate attempt to pass temperature checking that might have otherwise prevented them from coming to work. These are classic security risk management questions that would need to be considered in assessing the strengths and weaknesses of any system. Technology can and does fail all the time and the security of any system is only as strong as its weakest link, but even if you could address all of these issues, there is still a fundamental question of: Does this technology actually work? And given the potential size of the market for such things, it’s a one-billion-dollar-question. I spoke recently to John Honovich of iPVM.com.

iPVM.com is a world leading source of video surveillance information and they conduct independent reviews of surveillance systems and technology providing honest, objective information to buyers and security management. Unlike many in the security press, iPVM is a subscription-based service and so by not relying on advertising revenue, they’re able to be highly critical of surveillance vendors when they have it coming and many of them do. John and I spoke at length about the work iPVM have already done, testing vendors’ own claims in respect to fever detecting cameras. And I asked him, “Does this technology even work?”

John: The Washington Post ran an article recently saying they really don’t work and this created a bit of a firestorm because it beggared the question of what is it for something to work? The biggest issue is, is the expectations that have been set by companies selling these systems and marketing them have been terrible. The way these systems are marketed, they work poorly at actually detecting fevers. So the marketing show groups of people walking at the same time. It shows people with hats and with glasses and bangs covering their foreheads and people walking at angles, that in our testing we’ve done already five systems, including Dahua, Hikvision, Sunell that causes major problems with detecting people who actually have fevers. So that’s the biggest issue, that if you take the advertising, the marketing, and what the sales people are saying literally, which lots of buyers do, you’re going to have lots of problems with it working and with lots of people with fevers being missed.

 

Daniel: So the problem is false negatives that is not detecting people who potentially have an elevated temperature?



John: Yes. Now there’s the other risk in terms of the false positive. Now on your side of the globe as things are getting colder, it’s probably not a risk right now. On our side, the Northern Hemisphere, as temperatures get warmer, like 30°C, 32°C, etcetera, you are in much more risk. If someone walks in basically from using Fahrenheit, at least on our side, with 95°F out, if you’re walking outside in a 95°F to 100°F temperature and you walk inside, the risk would basically you having an elevated temperature just because you’ve been outside is very very high. So if the weather is 70°C degrees or like 25°C, you still have the risks of basically being missed because you keep with hats and hair and glasses and things like that, but then you also have the risk when it gets really warm that you’re going to basically get a false positives. But on the flip side, if it’s really cold and you walk in from when it’s 5°C, you have the risk basically that you may have a fever and you still might be missed because your forehead temperature is cold because it is cold outside and you just walked inside and right your forehead is still relatively cold from being out in the cold.

 

Daniel: If you’ve traveled through Asia any time of the last decade, you will see this sort of technology deployed at boarding gates and so forth and I imagine an airport is a much more controlled environment. They can control the climate, they control the numbers of crowds and the way those crowds behave, so that would seem like an ideal environment to deploy this as opposed to any other organisation where you just have a front door or a gate or something like this. The systems that are in use at airports, are they better or worse or do they work or they are also unreliable?

John: So there’s not many studies that we found so far. We found EU study for basically something between SARS and H1N1 and it wasn’t entirely encouraging from what we saw at the studies that very few people were detected with fevers, like a minuscule percentage, but we haven’t studied those studies enough and there’s not enough nor enough recent ones for us to be certain about airports. However to your point, some of the things about airports being indoors and that people have been indoors for some time and that you’re controlling people anyway, right? Because you’re in sort of a constrained environment where you expect to be searched, stopped, etcetera. Some of the infrastructure and logistics of airports lend itself better to do these types of screening.

To the contrary, take an example, we saw a case in Vancouver in a liquor store that had a Hikvision camera and whether it’s Hikvision or anyone’s camera, you would still have this issue. The camera was pointed at the front door entrance to the outside. So this is one of the worst case scenarios for having a fever camera or temperature camera. Every time someone opens a door, you’re either getting the heat or the cold from the outside. The person’s immediately being checked coming from the outside. So there are issues like that. And of course this is going from the sort of rare domain of airports where they’re making these sort of long-term planned out decisions to someone saying, “Oh my God, I want to open my liquor store,” and you’re going to throw one of these fever cameras hoping that you’re going to detect people who have fevers coming into your liquor store.

Daniel: So you’ve effectively got a perfect storm. We have an uneducated end-user who has no understanding of this and they just want to get their business moving. You have an uneducated integrator or installer who is traditionally used to just hanging cameras on a wall that just work and they are in-turn being misled by vendors who are saying that these products can do amazing things that are really beyond the realms of current technology. So you’ve really got the blind leading the blind selling to the blind, don’t you?

John: Oh I think in some ways it’s worse than that. It’s not that I think people are blind, it’s that people are in panics. So if you look at it from the seller’s, a lot of the sellers have faced dramatic drops in their revenue. So the sellers are panicking. Do I need to lay people off? I’m going to lose my business? The buyers in this case, like the liquor store. I don’t know that liquor store, but I would imagine that liquor store is concerned about people coming into their store and so they’re like, “Hey, how do I keep my liquor store open? Well, maybe I’ll try out this fever camera and I’ll pay the $10- $15- $20,000 these things go for to hopefully get my store back and running.” So I think that’s the other element. There’s been a lot of panic on multiple sides here that sort of driving this initially.

Daniel: So they are saying it as a cheap gamble really that look, it’s $10- $15,000. It may work, it may not work. But if that gets me in business that’s probably better than doing nothing. And so that’s the way they are seeing it?

John: Basically. I don’t know for a lot of these business. I wouldn’t necessarily call it cheap, but I think worse is you say like, “Hey, if my store shut down for a month, I’m going to lose $50,000 in revenue or a $100,000 revenue. So I’ll spend this $20,000 and if that makes my store open, well, then it’s worth it. So I think that’s some of the mental calculations going on to help justify it basically on the user side.

Daniel: Obviously security companies are also worried about the future of their business. So some of them have sort of pivoted into this space despite historically having no particular experience with it and there again, just relying on vendor claims, which may or may not be accurate but what is the size of this market?

John: We estimated that it’s likely to be billions of dollars this year, in US dollars. If you look at it in years past with something like tens of millions, maybe a hundred million dollars at most, as we both observe that it’s primarily with an airport type thing, right? It’s a very large-scale transit point and it has moved basically from airports to all over the place. Some people think this is the new normal and you’re going to measure temperatures as forever. But also some people think people are going to wear face masks forever. I’m less of the belief that you would do it forever for sort of more than a few years. The world will come back. It came back from the Spanish flu and many other different pandemics over the years. So I don’t see basically the white people in 2025 or still going to be using fever cameras to go to a liquor store. I think it’s a matter of time before the market crashes. But of course whether it crashes in 3 months or a year or three years makes a big difference in terms of how much money the sellers will make. For now, it’s clearly the hottest growing thing, that’s the video surveillance in almost 20 years.

 

Daniel: Yes, the last one that I saw was probably analytics 20 years ago and 20 years ago analytics didn’t work. But you had vendors telling people that “we just put this camera here and it will tell you when a person jumps the fence” or in a recently post 9/11 world when a suspicious object is going to get left lying around by terrorists.

John: There’s something really importantly different about the analytics case. Because in the analytics case it’s very easy to see that the analytics don’t work because the wind or a headlight or dust or rain would set off the analytic alert and you’d be like, “Well, it never seems to get people jumping over the fence and that’s a rare thing but it gets all these other types of things”. 

What’s really problematic about these fever cameras is that if you look at basically how a security installer would normally go about it, you would walk up to the camera, it says you have–I’ll do it in Celsius–it’s 36.8°C, right? It’s pretty good, right? That’s what I am. But the reality is if the opposite of analytics. Where analytics was really hard to make it ever work… I have made this joke a number of times – I could make a fever camera that was basically just a colour camera face detector with a random number generator and I would just basically say, “Hey, pick a number between 36.6°C and 37.1°C and just randomly pick from there.

Daniel: And people come to you and say, “Please take our money.”

John: People would be like, “Wow, this is amazing! How did you get so close? You guess 36.8°C and I’m 37.0°C. How did you get so close?” It’s like, well, it’s because 90-plus percent of people are within 1°C literally temperature range. So it isn’t that hard basically to essentially say everyone is sort of roughly 37°C because everyone is roughly 37°C. In temperatures, nobody has a 40°C normal temperature or 42°C or 33°C . If you had a 33°C body temperature, you would be on the verge of death. That is what is really fascinating about these fever things, that as long as a system throws out a number that’s close to what statistically average it looks super accurate.

Daniel: They fail quietly. So as opposed to say analytics where you knew you had a problem or any security system when you suffer loss in the security system didn’t pick up on it, you know you had a problem. Whereas this, you could have a massively infected person who could walk in passing infection to every surface and every individual in a building, the system wouldn’t go off and you would actually never know because they could come and leave without a trace.

John: There’s definitely that risk. The other thing I see in terms of–and I think that was one of the things we talked about–how do you test it is that the water bottle thing. Have you seen the water bottle thing on when people do marketing?

Daniel: Where they measuring the temperature of water bottles? Yes, I have.

John: The big thing with the water bottles is you take hot water and you put in the water bottle. You put it on your forehead, then you walk through and it’s like, wow. You got a 40°C degrees temperature reading. One thing we realize in our testing was that the water bottle thing isn’t fundamentally a bad test, but you need to control it to a very specific temperature like in Fahrenheit. When I’ve seen these demos, it’s like people have a hundred twelve degrees temperature. If you have a hundred twelve degree temperature, you’re not going to get there you’ll be dead at that point. Would you really care about this people at a hundred degrees, 101°C , 102°C ? So when you do that, that’s where you see the bigger issues and that’s more realistic because the likelihood of someone coming in into your facility with a 106°F temperature is really low. That is basically—

Daniel: That’s apocalyptic. There’s going to be other subtle clues that that person has a fever. John, I’m by no means an expert in thermography but I understand emissivity which is how much infrared a particular surface emits. So for example, a black carbon steel frypan might emit a certain level, whereas a stainless steel frypan would emit something else. So those two things that exactly the same temperature will read differently on an infrared thermometer that you use in the kitchen. So it would seem to me also that skin has an emissivity that perhaps a water bottle doesn’t, that clothing doesn’t, and non-contact thermometer. So the things that get pointed at your forehead have been calibrated to adjust for skin emissivity.

So for example, a lot of people very early in coronavirus might take their infrared thermometer that they used for testing their air conditioning or their electrical switch boards or their barbecue and then they quickly put that in the reception to measure the temperatures of people and of course they freaked out because those temperatures were quite a few degrees lower than where they should have been. So it would seem to me that cameras have this similar challenge that they have to meet. So even the water bottle test, knowing that a clear plastic water bottle is going to transmit heat and light very differently to skin seems to me to be the field of sort of circus tricks rather than actual science.

John: Yes. I mean, you could really heat up your forehead. So what we found basically is that Sunell and Hikvision only look for temperature, read temperatures on the forehead. What we’ve done with our testing is that you can heat up a forehead and you can control it roughly by using basically a handheld thermometer to double-check. That you can get it to a 101°F or 102°F etcetera. You don’t want to get to 105°F or 106°F because again, that’s not terribly realistic. It’s going to be extremely rare that someone’s walking into basically any facility with a 105°F and you’re not going to notice. 

 

 

 

Daniel: They’re not going to be walking!

John: So what we did is we heat it up to forehead with the water bottle, but then took the water bottle off and have someone walk through/ What we found was, that especially when you had hair or hats or walking through, that many of them were missed and the system came in a low. So what we found was that the systems that we’ve tested so far–Dahua, Hikvision, Sunell–tend to normalise readings. Meaning that lower readings or bad readings tended to become more normalized like closer to 37°C and that higher readings tend to come in lower. And that’s a risk of course. If you really want to know if someone basically has “fever.” You definitely want to make sure that you have if someone has a 37.5°C or 37.8°C temperature, you want to make sure you get it and it doesn’t get read lower as like 37.1°C, for example.

Daniel: So do some of these systems auto adjust and re-calibrate themselves over time?

John: No, they don’t auto re-calibrate. I mean, you need to make sure that you are calibrating them yourself. There’s no magic. Even basic when you calibrate these systems, you have to manually input what the ambient temperature is. So if you make a mistake, let’s say you put it in that it’s 70 and it’s actually 65 that will throw off your readings.

Daniel: Right. So for example, if people were deploying this technology outside, for example at a stadium or a major venue or a vehicle or entry or something like that. It must fail.

John: It won’t “must fail” because most people don’t have fevers. So the problem is anytime you’re basically doing this type of fever thing, it’s like a needle in a haystack type of application. Let’s simplify it, instead of saying the temperature, we are going to say fever or not fever, we are going to make it binary, okay? And I say, “Not a fever, not a fever, not a fever, not a fever, not a fever,” right? And then we had someone do the ground truth and naturally tested them with an oral thermometer, 99 times out of 100 or 999 times out of a 1000, I would be right because people don’t have fever. Fever is a very rare condition. So that’s basically the challenge there. The issue when you’re going to do it outdoors is that you’re going to have variations in temperature whether the position of the sun, clouds will make it slightly hotter or colder.

 

Daniel: People’s clothing as well which would warm their body up I imagine.

John: Also, how long people have been outside. If you think about it, like people talk about doing this in a casino. It was like Las Vegas at least in the U.S. is kind of the big mental image. The issue is if you’re walking outside and go indoors and immediately you are checked, you’re going to have a different surface temperature than someone who got out of an Uber or out of a taxi and was in basically air conditioning and just walk straight into the casino. So those are some of the challenges that you are getting into when you’re trying to do these types of things outdoors. And so you can compensate.

We’ve seen things like, you can set the alarm threshold higher. Like if you don’t really trust your system, you can say, “Well, you know what, because people are coming in from outdoors, we’re getting too many people that are being stopped because whatever they were just walking outside and it was warm. So instead of making the alert temperature be at 37.5°C let’s move it to 38°C. Let’s move it to 38.2°C, let’s move it to 38.4°C. So you gain this classic trade-off between false negatives and false positive. If you keep on moving it up, well, great. You’re going to have less false positive. Though the flip side problem you have is that other people basically who have fevers who weren’t air conditioning or put water on their face or whatever it was, those people are going to be missed and go through. So it’s a really hard thing to do.

There are standards to this. There’s the IC Global Standards, the US FDA is done testing and the technology itself, at least the underlying technology of demography is fairly mature. But the understanding when you looking at these studies and looking at the standards and government regulations is that you need to do this in a controlled environment. And the really big issue that you get here is that academia and governments and standards all say, “Hey, you’ve got to basically do this in this very controlled matter.” And then the marketing of the security company’s is like, “Ha-ha-ha. You can wear a helmet and you can walk through with six other people and we’re going to do this accurately.” And so that’s the problem if the vendors were just say like, “Hey, we understand what the global standards are. We understand what the studies say. We’re going to follow this.” Most of the problems would go away.

Daniel: So John, obviously researching on Coronavirus is breaking in real time, but based on what we know today up to a quarter of people who are infected can exhibit no symptoms whatsoever so they can be walking around as a completely asymptomatic carrier. So we already know that even of the needle in the haystack, the people who are infected, a quarter of those individuals may not show any symptoms. What are some of the outrageous claims that you’ve seen vendors making?

John: I’ll be giving some funny example. When we did our testing, so I tell our team when we do testing, “I want you to go to the company’s website and I want you to test what they claim. Literally. Read the website, look at the pictures. If they say that they can detect a dog wearing a tuxedo, then you test with the dog with a tuxedo. And if they say that basically they can do fever detection with someone wearing a helmet or hat and glasses or three people side by side, then that’s when we test.

Daniel: Right.

John: So we do that. And then we went to the manufacturers. Two of the manufacturers happen with and then they came back and the engineering team said like, “Oh, no, no, we don’t recommend you do that. You shouldn’t wear hats or you shouldn’t do this.” And then I said to myself, “Wait, have you seen your own company’s marketing? Your own company’s marketing is filled with examples of this.” A lot of these companies engineering team fundamentally knows how these things work. Or don’t work right? In this instance, but the sales and marketing team and whoever that works out, I don’t know how they got there. But the story that the sales and marketing teams are saying diverges from what the engineering team actually understands.

Daniel: It seems that in some respects this is similar to any other technology product that they really only have two opportunities to stand out. One is in terms of their capability and what their systems can do and the other is price and they’re probably all sort of competing on both fronts. So they’re all trying to say, “Well our camera technology is better than someone else.” Or “Our technologies is just cheaper.” One of the things that I found interesting is the amount of this technology that’s coming out of China. So you mentioned two companies that are doing that. Now, on the one hand, Asian countries obviously have a lot more experience than many other Western countries do in dealing with pandemics. They’ve been dealing with it for years. There’s a reason that Sydney airport never did temperature detection. But if you’ve got off a plane in Hong Kong 5 to 10 years ago, they did. So on the one hand, they’re probably more familiar with it, but on the other hand, there have been some major issues with Chinese security technology and I know, for example, the number of vendors have been outright banned at US government facilities because of fears of the Communist Chinese government actually using that as a spying technology. How are you finding that government consumers is in the broader market of managing this divide?

John: Two things. One, that indeed experience thing with China companies is not accurate in this context. If you look at Dahua and Hikvision, from everything that we’ve researched they’ve rushed out these solutions in the last couple of months. They were not doing human body skin temperature detection prior to the coronavirus calm down. Sunell was but Sunell was doing this in schools and Sunell has their own issues because they’ve got this relatively speaking super wide field of view which creates other problems when you’re trying to accurately detect fevers in this type of environment. So I don’t think that the China manufacturers, especially the ones that marketing in security for sure have like this advanced expertise in fever detection.

For the government issue, Dahua and Hikvision are banned for US government use in August of this year, just three months from now. They’re scheduled to be banned for federal government for broader use as well as the “blacklisting” clause for government contractor. So there’s been a lot of panic in general in terms of what to do with these systems. Though it still seems for the most part more critical infrastructure is going with FLIR but there’s the famous example of Amazon even in the USA using Dahua which resulted in the Reuters investigation, a lot of basically questions about that.

Daniel: So John, you mentioned FLIR. For those who aren’t aware, FLIR is an absolute market leader in thermal cameras and thermal infrared sensing and so forth. And I mean that traditionally has been for security applications. So spotting a person through the bushes or through the fog in the freezing cold of night where traditional cameras rather sensing may work and they’re very good at that. They’ve got a lot of great stuff and they are also strong in the sort of scientific sensing an emergency sensing field. But I mean, they made a comment on their website where they actually state that there is no way to thermally detect an infected individual who doesn’t have an elevated skin temperature and only a licensed Medical Professional can determine if a hot individual is experiencing an abnormal medical condition. In other words on their website, they have this massive disclaimer saying their stuff doesn’t work for detecting coronavirus.

John: Yes, that’s because FLIR is an established us company that understands how US Food and Drug Administration regulations work. Though you’ll see similar disclaimers in many fever camera, offerings saying that this does not detect coronavirus, etcetera.

Daniel: Now, to FLIR’s credit though, they’ve been very upfront about it and they’ve basically said this is the wrong tool for that solution, but their technology is now being leveraged by other people who were sort of piggybacking off the prestige of their brand.

John: They are marketing elevated skin temperature detection and they are saying that per FDA regulations or guidelines that when it’s used with FDA-regulated clinical thermometer that it can be used for this. They also are FDA cleared to do this. The more interesting dynamic that I see between FLIR and many of the new entrants is that FLIR has been consistently very conservative in how they recommend this to be used. They literally talk about people to stop take off their glasses. This is much different than most of the other companies that are marketing this as there are saying like, “Hey, don’t worry about glasses or hats or hair or anything. You just sort of walked by and we got this.” From everything we’ve seen everything, even from the FDA studies and from other academic studies, we haven’t seen any studies that say, “Put three people side-by-side wearing helmets and hats and glasses and whatever and you’re going to get accurate results.” That doesn’t exist from everything that we’ve seen. The engineering teams know this as well. When you look at Sunell, they have marketing where they’re doing a 12, 24, 40 people at one time. But when you read, the installation instructions are very clear that they want it one at a time.

Daniel: And so you have different applications. For example that you have, a liquor store would have a very different use case than say a stadium or a nursing home or a hospital or a hotel. But even those facilities themselves have different use cases. So for example, a hotel might have a predictable number of visitors coming in through a door at any particular time and that may fall within the manufacturer recommendations, but they would also have staff shift changes where you might have 400 people coming and going at the same time which would completely throw that out of whack. People are sort of looking for a one-size-fits-all and it just simply doesn’t exist, does it?

John: The one-size-fits-all doesn’t per se exist. The issue is that we’ve heard this from most of users as well, is that it has to be high throughput but the real answer if it has to be high throughput then you’re going to miss lots of people with fevers, that’s the reality. You have to take your pick. The problem is the sales and marketing people who are saying like, “Yeah, we understand it has to be high throughput and we have a high throughput application.” We’ve seen things where it says 30 people per second, 18 people per second, X thousands per minute or whatever and that’s the problem–to say that you can do it accurately and do this high throughput–that’s the problem.

Daniel: It’s one of these interesting risk challenges. I mean, if you looked at site traditional metal detection, if you said to aviation or an airport or a stadium that these metal detectors will only let through one percent of people carrying a fully automatic weapon, they would freak out.

John: Right, yes.

Daniel: They would never buy into it. Because they would apply that and risk management decide, “Well, that’s not a risk that we’re prepared to wear.” But in this case, that particular percentage maybe comfortable, of course, unlike a person with an automatic weapon who then decides to use it. This person is not going to reveal themselves until everything is long since gone.

John: I think your analogy is good. Basically, like a metal detector let’s more people get through with guns. The issue right now is this is—so we’re really only in month 4 when we talk about it in terms of outside of China where we are only in month 2 but I can remember all the way back to April. April was initially like a panic. I think we’re already seeing now is that you’re getting people more informed and thinking more rationally about it month by month. So I expect that to continue and the people will have a greater sophistication, understanding of the subtleties of using this type of technology.

It’s really hard if you are a Security Director and you need to make this decision within what mid-March basically when it went and sort of global. Between mid-March and April 15th, while people are being locked at home. That’s a really sort of difficult decision to make without understanding the technology, with this panic: am I going to die? Is my family going to die? Is my business going to go out of business? I think that was super difficult, especially during the first month. I’m quite confident that people will continuously act more rationally about these technologies. As time goes by and they get more of an understanding of what the real trade-offs of using these things are.

Daniel: So John, if you had a Security Manager or government procurement or someone who had been told we need this, this is the way to go or perhaps, this is just one of the tools we’re going to add to our arsenal, what resources would you point them toward in being able to identify fact from fiction and things which work well, things which work so-so and things which just simply don’t work?

John: iPVM is literally the only site in the world that is actually doing these types of thorough testing of various products. We’ve been the world leader. We’ve been quoted in publications throughout the world for what we’re doing. We have four now. We have like five people working full-time just testing these systems. We’ve literally shut down testing anything else right now and all we do is basically test these sort of fever cameras and systems because there’s an immense demand to get help base to figure out how do these things work. Well, even right now, we’re testing the kiosk system, the tablet ones that you walk up to. They’re less expensive and we’re trying to figure out what are the various issues with that. So quite honestly iPVM.

Daniel: That was John Honovich of iPVM.com. My name is Daniel Lewkovitz from Calamity. And if you care about your security and would like to talk to someone about a security system that works including installation of CCTV as well as 24/7 monitoring of alarms and video, you can ring Calamity 24/7 on 1300 300 24 7 or visit us at calamity.com.au

[end]