I gather you are joking, but for a serious answer:
There is no vehicle model # information, we asked Verkada if they have vehicle focused analytics (e.g. car/truck/SUV, color, model) on their roadmap. I will update here when we find out.
"with IR on, unsurprisingly given the lack of color captured, identified all objects as white or gray."
Color analysis should have been made unavailable when using IR - and they should've added a pop-up message telling the user why only an idiot might try and use it when IR is on.
but this analytic is a forensic tool (i.e. using recorded video) vs using live video and databases to provide exclusion via black list and/or inclusion via white list.
I agree that end users overall find more value in facial alerting rather than searching but it is easy / less expensive to offer searching, so it is a prudent first step.
I could imagine Verkada eventually offering real-time as an upsell / premium feature.
Seems to me it’s less of an ethical issue, since the majority of us agree that your chromosomes choose your gender, not your mood. Nothing wrong with being non-binary or whatever the term you’re most comfortable with is, but until we can add some more options to the forms, and educate people out of being weirded out by it, it just is what it is.
I see your point, but I'm talking about the ethical dilemma here. The ethical questions surrounding the "why" when you're interested in ethnicity are far greater than when you're simply trying to get male/female identification for various reasons.
For example, what China is doing with analytics to track ethnic minority activity is clearly indefensible, but someone trying to say whether a suspect is male or female so the police can track them down is not even close to being the same thing.
Others here have made valid points about using this type of analytic in a retail setting in order to maximize product exposure to a specific demographic, which is also a use case with few ethical issues, at least not as many as the race thing. To avoid this, you can simply not shop in stores with video surveillance. Use Amazon I guess...but of course, they track and ID you in their own ways.
"its called Marketing Data, it has a ton of value in BI for Retail"
i hear this all the time from the sales side of our industry... and I am still pretty suspect of the actual value of retail analytics.
Just because you can capture data (heat mapping, people counting, etc) does not mean that this data automatically translates to actionable BI.
Retail Merchandising is already (and has been for a long time) an actual scientific thing. You can get a Masters degree in this field.
Merchants (especially large, multi-location) do not stock and position products based on guessing. i.e. any BI produced by cameras with retail analytics is largely already known.
it is my position that the 'valuable data' captured by cameras (with analytics) is marginal at best - yet the cost is not.
Correct, though Verkada is not marketing this as a retail analytics solution. Of course, they might at some point, but for now, unlike typical retail analytics offerings, this is a free extra.
PPF is what? Pixels Per Frame? While some people use acronyms to quicken their reading and writing, it becomes a stumbling block to a wider readership. I suggest always defining an acronym at it's first instance in any article. This way you reach a broader audience and help educate those less informed with current acronyms.
That term is so fundamental to video surveillance we do expect it to be known or someone can search for it on IPVM. For sure, some people will not but because this is a specialist site for video surveillance, we tend not to link to or explain fundamental terms. We do, of course, have reference / training material like the guide above.
Is there a page, e.g. glossary of IPVM (and I cannot recall what IPVM stands for, but I know it when I see it) terms, that has all your fundamental terms you expect of your readers so I do not have to wade through reference/training material? I subscribe to your site because it's on the cutting edge and the topic interests me; I do not deal with this area day in and day out. It would be helpful to have a single page I can refer to when acronyms are resorted to if the policy is not to define an acronym at its first usage.
Nice page re: Pixels Per Foot and I do not recall reading it, though I may have years ago.
while I do not disagree with your premise, if you don't know what PPF is, how does the information in the OP (original post) make any sense at all to you?
this was a serious question - maybe delivered in an unnecessarily snarky fashion.
understanding fundamentals is required before you can really understand solutions... like why choose this camera over that one for my particular environment.
I welcome new members who want to learn the fundamentals in our industry - like Mr. Poole appears to be doing...
I would suggest that anyone in his position enroll in the IPVM (IP VIdeo Market) Camera Course - or at least study the available IPVM documentation that is used in this course.
It is, by far, the most comprehensive (and up to date) material you can find anywhere.
Poole sees PPF, and wonders what that means. Gets an idea to make things more user friendly. I'm guessing that an automated parser can process the text and add a tool-tip on common terms that may or may not be immediately recognized by newcomers.
Anon: If you don't know PPF you probably don't understand anything about any of this, please enroll in an IPVM course.
I know that IPVM is a forum for the exclusive l33t IP video gurus, but maybe with a little effort mere humans might get an improved experience without signing up for a course.
"I'm guessing that an automated parser can process the text and add a tool-tip on common terms that may or may not be immediately recognized by newcomers."
personally, I loathe info links on 'common' terms within paragraphs - as I generally use my mouse to follow what I am reading and these links tend to hijack me into windows with links to things I don't need.
Instead, maybe a post pinned to the top that lists all known and made-up initialisms/acronyms commonly used?
"maybe with a little effort mere humans might get an improved experience without signing up for a course."
right. this is why I suggested an alternative - study the available documentation on IPVM which covers the majority of the content used in their courses.
A/B testing would offer some insight into the best way to handle this sort of thing.
As a side note, some sites are able to handle imperial to metric conversion as an option so as to spare members of more advanced civilizations from the unbearable agony of enduring exposure to these vulgar medieval units.
I'm a little surprised at how well they performed in the test. I was sent a few of their cameras for testing six months ago and thought that there image quality was below industry standards. Their inability to adapt to scene changes, like opening a bay door, immediately disqualified their cameras in many of the markets we service. Additionally, ghosting was an issue across the board.
That being the case, when i heard of their FR analytics i thought there image quality would take away from the analytics reliability.
Has IPVM published any articles on how image quality effects analytics? Maybe I've misunderstood how algorithms work. I always thought it was a mix of Pixels on target or PPF and image quality.
We did a test of their video quality (here) and what we found agrees with you. They were reddish during the day, noisy and artifacty at night, and had trails in WDR scenes.
More generally, image quality impact on face rec is essentially going to come down to whether enough resolution is delivered at good even exposure and contrast. If any of these are a problem, the face will appear blurry and recognition will suffer.
Also, to be clear, in our tests Verkada's facial detection was reliable, but at short ranges, only 8-9' from the camera, which was ~130 PPF, which is pretty high. That is not to say that recognition works at that level, as the camera definitely did not catalog every face, but we haven't quantified that specifically.
Thanks for the clarifications. May I ask how they performed in dark environments? It’s my understanding that poor light can cause a decrease of PPF from 20-50%.
I guess I’m happy to see them send product to an integrator, I keep seeing them in my customer’s offices. In my area we aren’t their channel. So they can pound sand.
"I guess I’m happy to see them send product to an integrator"
while I love your pound sand position for it's strength, why bother even showing up at customer sites to 'compete' against a a manufacturer playing outside the traditional channel?
you can always email your pound sand position without wasting your time showing up at sites... unless maybe you are just paying offloading your lunch bill ?
Sorry, I misread Undisclosed for Integrator. I don’t expect you to agree with my POV as any other part of the channel, whatever it is that you do.
We show up at all “good” opportunities and we’ve managed to retain our customers that Cisco and others have shipped product in to. My turn- I don’t understand this part of your comment, we compete every day for new and to retain existing customers.
No, you replied to a UD Integrator - it just wasn't me, but I then decided to weigh in on your conversation. please pardon my impertinent comment.
I agree that you should show up for any potentially rewarding opportunity.
but having at least a general understanding of where potential customers value judgements reside on the food chain is integral (imo) in avoiding obviously wasted efforts just showing up at any 'event' hosted by whomever.
I guess? I just don’t see how the local Cisco RSM shipping in some cameras to a customer means its curtain for anyone else. I’m searching for something in your comments that’s relative to my comments. No, no one is going to walk out the moment a Meraki camera shows up, sorry? If for some reason I get the intel that the customer is a price only, Amazon buying tote of guy or gal- I’m out.
i am not trying to be argumentative... believe that or not.
i refer you back to my original comment that was based on your own words:
"I guess I’m happy to see them send product to an integrator"
i still do not understand why you are happy about end run moves around the channel - when it's clear that you are an industry professional that understands the strength of that channel.
i discovered they have a very large team of unqualified surveillance specialist. Neither the sales person I spoke to nor her manager understood basic concepts such as IRE or even WDR.
There default sales pitch relies on there ability to raise capital not image quality. I thought it was odd approach, however with the success of Shark Tank, maybe old security dogs like you and I are doing it the wrong way.
Keep selling Sony, my friend!!
I’ll give you a call later in the week to catch-up.
Yeah, this guy’s POV is exactly that of NI groups and the Cisco chain. I’ve dealt with them for 20 years and wish I could make the money they do. I wouldn’t be surprised if I knew the guy commenting, he reminds me of one in particular Cisco regional guy... lol
Our industry is still specialized enough for high demand customers to need us. Will a plug it in and web connect camera replace what 80% or more of the security resellers do out there? Maybe. Knock out all the Hik and Dahua NVR’s you can get fellas. But we won’t install it for them. And anyone else who uses their DCJS or related licensing to support it is a fool. They’ll do it on their own.
Saw you had a new driver in the household, good luck and talk to you soon!
I received my test cameras, one of each of the models they offer. They've been running in my test environment for about a week, and I'm pretty impressed with them so far.
I had some issues w/ video recall at first, but once I worked with their tech support to push a firmware update it resolved itself right away.
Installation was ridiculously easy, and setup was basically just connecting it to PoE and letting it do its thing. I don't anticipate using something like this at any of my Tier 1 sites since they have to have to communicate over the internet to Verkada, but I'm seriously considering using these at lower tier facilities that aren't considered critical infrastructure. I'm thinking administrative areas, customer service, and anywhere else the people analytics would be helpful.
I ran some cost benefit analysis and it shows my maintenance costs plummeting if I implement around 350-400 of these as replacements for our existing IP cams in admin areas. No more NVRs, no more licensing, and no more constant stressing about whether or not the incident I need to review had been overwritten since the department waited two months to ask me if we have this obscure incident at a remote site on video. 120 days of onboard recorded video, without the risk of an NVR failure taking the whole site down is a big win.
In any case, I'm impressed, and that's not easy to do when it comes to this sort of thing.
Hi Michael, in that case, we'd probably lose the video, similar to any other camera with onboard storage that isn't removable. If the camera isn't completely dead and I can still communicate with it, I suppose I could archive all of the video to the Verkada cloud so it's accessible.
If I do order like 400 of these for production, I would likely order another 10 of each model to keep on hand for quick replacement while the broken one is getting repaired/replaced. Loss of video is one of the other reasons I wouldn't use it at our Tier 1 sites. We go through a lot of effort at Tier 1's to build redundancy. As long as I'm informed of the incident in time, I have a few places I could pull it from. In those locations, we keep some video onboard the camera, we keep at least 30 days on the NVR, and we have even more available in external storage using the Veracity Coldstore device which has built in redundancy as well.
Hey Michael, how do you guys get the cameras to the web? We spend a ton of time in the utility space and have never seen any device that touches the SCADA network in any capacity routed out to the internet. Do you guys have separate physical networks at your subs?
Hi Jason, we have separate networks for the business, SCADA, and Security. Tier 1 facilities have their own network as well. Even with that though, there’s no chance of me letting anything that talks to the internet anywhere near a tier 1...
After letting the Verkada cameras run for several weeks, we've updated the section on their face search capabilities above:
First, subjects were frequently added to the database as multiple unique faces instead of being merged into one. For example, in the image below, the same person is shown as 10 unique subjects (highlighted in red).
Note that these duplicates does not return the same video results. Event clips from different times are shown for each.
Additionally, some subjects were incorrectly recognized. For example, clicking on one person returns detection results for 5 different people. This one person was recognized as almost everyone in our office.
"This one person was recognized as almost everyone in our office."
except for those with beards
Agree
Disagree
Informative
Unhelpful
Funny
Create New Topic
Subscribe to IPVM Research to read the full report.
Why do I need to subscribe?
The IPVM Research Service includes products tests and shootouts plus competitive and financial analysis, helping decision-makers better evaluate purchasing, partnering, developing, and/or competing against companies in physical security.
Comments (55)
Undisclosed Manufacturer #1
"Dude looks like a lady" -Aerosmith
Create New Topic
Undisclosed #2
can you list vehicle model# and gender? curious :)
Create New Topic
Sean Patton
I gather you are joking, but for a serious answer:
There is no vehicle model # information, we asked Verkada if they have vehicle focused analytics (e.g. car/truck/SUV, color, model) on their roadmap. I will update here when we find out.
Create New Topic
Undisclosed End User #3
Good report IPVM.
Anyone else turned off by enterprise security being referred to as a game ?
Create New Topic
Undisclosed #4
Color analysis should have been made unavailable when using IR - and they should've added a pop-up message telling the user why only an idiot might try and use it when IR is on.
Create New Topic
Undisclosed Manufacturer #5
From my experience, end users wants real-time facial rec. Who wants to find out 40 minutes later that there is someone on the blacklist inside?
Create New Topic
Jonathan Strauss
But can it identify a hot dog?
Create New Topic
Michael Gonzalez
08/17/19 01:43am
Hmmm, I have some test cameras on the way, this is good to know. I'll post some comments after I've tried to break/successfully broken them.
Create New Topic
Undisclosed Integrator #6
Real people struggle with gender analytics in this day and age anyway so I'll give them a pass on that one
Create New Topic
Undisclosed #2
why is gender analytics socially acceptable? shouldn’t it be treated the same as ethnic analytics?
Create New Topic
Undisclosed #4
i hear this all the time from the sales side of our industry... and I am still pretty suspect of the actual value of retail analytics.
Just because you can capture data (heat mapping, people counting, etc) does not mean that this data automatically translates to actionable BI.
Retail Merchandising is already (and has been for a long time) an actual scientific thing. You can get a Masters degree in this field.
Merchants (especially large, multi-location) do not stock and position products based on guessing. i.e. any BI produced by cameras with retail analytics is largely already known.
it is my position that the 'valuable data' captured by cameras (with analytics) is marginal at best - yet the cost is not.
Create New Topic
John Poole
PPF is what? Pixels Per Frame? While some people use acronyms to quicken their reading and writing, it becomes a stumbling block to a wider readership. I suggest always defining an acronym at it's first instance in any article. This way you reach a broader audience and help educate those less informed with current acronyms.
Create New Topic
Undisclosed Integrator #9
I'm a little surprised at how well they performed in the test. I was sent a few of their cameras for testing six months ago and thought that there image quality was below industry standards. Their inability to adapt to scene changes, like opening a bay door, immediately disqualified their cameras in many of the markets we service. Additionally, ghosting was an issue across the board.
That being the case, when i heard of their FR analytics i thought there image quality would take away from the analytics reliability.
Has IPVM published any articles on how image quality effects analytics? Maybe I've misunderstood how algorithms work. I always thought it was a mix of Pixels on target or PPF and image quality.
Create New Topic
Michael Gonzalez
09/05/19 07:08pm
I received my test cameras, one of each of the models they offer. They've been running in my test environment for about a week, and I'm pretty impressed with them so far.
I had some issues w/ video recall at first, but once I worked with their tech support to push a firmware update it resolved itself right away.
Installation was ridiculously easy, and setup was basically just connecting it to PoE and letting it do its thing. I don't anticipate using something like this at any of my Tier 1 sites since they have to have to communicate over the internet to Verkada, but I'm seriously considering using these at lower tier facilities that aren't considered critical infrastructure. I'm thinking administrative areas, customer service, and anywhere else the people analytics would be helpful.
I ran some cost benefit analysis and it shows my maintenance costs plummeting if I implement around 350-400 of these as replacements for our existing IP cams in admin areas. No more NVRs, no more licensing, and no more constant stressing about whether or not the incident I need to review had been overwritten since the department waited two months to ask me if we have this obscure incident at a remote site on video. 120 days of onboard recorded video, without the risk of an NVR failure taking the whole site down is a big win.
In any case, I'm impressed, and that's not easy to do when it comes to this sort of thing.
Create New Topic
Ethan Ace
After letting the Verkada cameras run for several weeks, we've updated the section on their face search capabilities above:
First, subjects were frequently added to the database as multiple unique faces instead of being merged into one. For example, in the image below, the same person is shown as 10 unique subjects (highlighted in red).
Note that these duplicates does not return the same video results. Event clips from different times are shown for each.
Additionally, some subjects were incorrectly recognized. For example, clicking on one person returns detection results for 5 different people. This one person was recognized as almost everyone in our office.
Create New Topic
Undisclosed #4
except for those with beards
Create New Topic