Don't Be Fooled By Hot Water Bottle Fever Camera Demos
By Ethan Ace and John Honovich, Published Aug 24, 2020, 09:48am EDTFever camera salesmen like to fool buyers (and themselves) with hot water bottle demos. It deceives and detracts from the real flaw - these systems miss real human elevated temperatures. Learn how / watch now:
What's clever is that many people see it and think "Wow, this thing works. No water bottle, normal temperature. Hot water bottle, alarm!"
Common Sales Tactic
The hot water bottle is the go-to move of fever camera marketers around the world such as Dahua, Hikvision, and Sunell, examples shown below.
Problem: Too Hot
The problem is simple, hot water can be really hot - 110°F, 120°F, etc. Just because a system can detect a 110°F bottle does not mean that it can accurately detect 100°F or 101°F, the real temperatures that humans would actually be with a fever.
Your system needs to be precise enough to distinguish between 98°F and 100°F, a much harder task than the fever camera salesman who wants to fool you with a water bottle of 110°F vs 98°F.
Solution Just Hot Enough
If you really want to see if a system works, you want to elevate your forehead temperature just hot enough to replicate 100°F or 101°F, the temperatures that actual humans with fevers have. A person with an actual 110°F body temperature is not going to be walking through anyone's entrance.
Here is how to do it:
Warm one's forehead with a rubber water bottle / pouch:
The aim is to heat one's forehead modestly. You want it to get to 100°F to 102°F to fairly emulate a 'fever'. This depends on how hot the water inside the pouch is and one can easily experiment with a few seconds to ten seconds to see the right duration to elevate a subject's forehead to the desired range.
Then walk in front of the camera or detector you are testing:
Immediately, have a second person come in with a calibrated hand-held IR thermometer and measure the subject's temperature.
Compare the temperature that the fever camera / detector produces to that of the handheld IR thermometer.
This way you can compare to ground truth (with the calibrated handheld IR thermometer) rather than simply walking around with a very hot water bottle affixed to one's forehead.
Don't be fooled, use this superior method to better detect accuracy issues. In our testing of 20+ 'fever' devices, we found most systems have issues detecting moderately elevated temperatures, tending to favor reporting back 'normal' / statistically average temperatures, which produces a false sense of safety and risk of infectious people passing inside one's facility.
2 reports cite this report:
Comments (14)
Don't be fooled.
Or better yet, stay far away from this trend. Don't sell "fever cameras" They will go out of style soon and then we will all look like fools for spending all this money needlessly.
I just don't understand this, as you have reported some of these systems work well if they are set up and operated correctly.
Why are they still being advertised in this obviously false way.
The argument about how effective they are is something else entirely :)
When I ran my tests I microwaved a bag of rice and held it to the subjects forehead for a minute or so, verify with handheld, then test.
Potayto/Potahto
Surprised what passes a proof of functionality these days.
Similar. In testing my unit, I put a wet towel in the microwave for 30 seconds, then held it to my wrist, then did the scan. When I held the towel up directly, I had a 160f fever :)
Great commentary!
What still amazes me is having to use an accurate ~$125 device to calibrate one’s forehead to be used by an inaccurate ~$5000 device! Go figure!
Sadly, the hot-water-bottle demonstrations are not just the domain of camera salesmen.
From the BBC (skip to 0:20 for the hot water demonstration):
From ITV News (skip to 1:30 for the hot water demonstration):
It's a shame that journalists at major, respected news outlets are repeating these hot-water-bottle demonstrations in their segments, which in turn only promotes further adoption of 'fever' camera systems.
It's a shame that journalists
The problem is that this market exploded so quickly so it was really hard to immediately understand what was happening. The hot water bottle demo 'works' at some very basic level (it can tell if the hot water bottle is 5 or 10 degrees warmer than a normal human). The problem is that these systems, to work for fevers (its intended purpose), needs to work at 1 or 2 degrees differential, which is the real practical problem for 'fever' cameras.
Great... now I know my recent internal testing was rubbish! This certainly explains the results -- boiling flesh but a temperature of 102.7 degrees. Back to the testing.
I can't see selling these without the explicit acknowledgement from the buyer that subjects with an elevated temperature are estimated (in the last thing I read, seems to be a moving target) to only represent about 25% of the possible infectious carriers of the virus. Other infectious persons can be prior to a fever stage, or asymptomatic.
All of these comments are why I stopped conversing with management on this topic. They aren't bringing it up and I'm not talking about it again unless they do. I do like the wall mounted IR thermometers and continue to explore them as an approach for temp checks (we are implementing weekly Covid tests for designated onsite personell anyway).
Hikvision hosted a webinar marketing its DS-K1T671TM screening terminal using a hot water bottle/mug for demonstration of the system being able to detect high temperatures on a forehead (compared to systems with a narrow scanning area):
There is no ground truth of the water bottle temperature in the Hikvision demo, and it could easily be 104 or higher.
