Anybody know of installations that has been done with IBM as VCA for large outdoor environment and perimeter detection with Milestone as VMS?
If you do can you share some information about it?
Thanks in advance
NOTICE: This comment has been moved to its own discussion: IBM Video Analytics Experience?
Ethan, thank you, I understand that.
I just wanted to read old discussions, that haven't been moved here. :)
Kirill, we nolonger use the LinkedIn group since opening discussions here. It's completely closed and inactive.
John, thanks a lot - that's all make sense.
May I ask you to give me access to linkedin group?
VCA Technology a subsidiary of UDP Technology licenses full feature video analytics integrated into NUUO, Luxriot, Digifort and many others. They provide to usual suite of features like object detect/classify, line crossings/perimeters, etc...
Intelligent video analytics are feasible, it's just a question of what types. Faces in a crowd, a bomb left in a busy train station, those are incredibly hard to do (related: see recent Facial Recognition Test Results Shared). Tracking a car across a city based on its size and color is another commonly fantasized but highly infeasible one.
License plate recognition, perimeter detection in unpopulated areas, to the contrary are two that can work well.
PSIM is fine for cities but it's not going to transform the city just because you have video, access control and a few other security systems on a map.
John, as I understand one of the main part of the "Smart City" product set should be some kind of PSIM. I've read 2 articles about it here:
And I can't say that these articles are optimistic. :)
So there's no applicable IVA for real city surveilance (only for specific applications) and no applicable central management system. ))
"Smart City" is basically a marketing term wrapping various security and surveillance technologies and then applying an outrageous price. I wouldn't say it's science fiction, but parts of their advanced technology claims are.
Hi Tiago, thank you for answer!
I'll take a look at these products.
you could take a look at Cognitec FaceVACS-VideoScan and Artec ID Broadway 3D.
The former we have performed some extensive tests regarding facial recognition. The results showed a satisfactory facial recognition of approximately 60%. That is, in 6 out of 10 attempts subjects of interest were recognized. It is important here to mention that many factors can influence the facial recognition rate, such as lighting conditions, subjects that do not look at the camera straight on, amongst others. With regard to assistance, it is worth mentioning here that Cognitec staff members were very helpful by providing us with tips about how to tune up our configuration.
In its turn, we have not evaluated the latter, but we have seen it in action. It looked very impressive, specially when (unsuccessfully) trying to fool the system, for instance, by holding up a picture or wearing dark sunglasses. They also claim to be able to recognize one person per second, or 60 faces per minute. However, we cannot confirm this claim because we have not tested it.
NOTICE: This comment has been moved to its own discussion: Facial Recognition Test Results Shared
John, here's new questions.
So based on your answers, may we say that "Smart City" concept is a science fiction also? What do you think about it and about vendors who claim that they have solution for this like IBM, NEC?
Keep the questions coming!
The crowded environment analytic scenario is fundamentally unsolvable.
For abandoned objects, the problem is that objects are continuously being obscured by people moving in front of them. Indeed, something could be blocked for long periods of time.
If you trigger an alarm too quickly (say an object has been stationary for a minute), you have tons of false positives and people hate the system. If you wait a long time (e.g., object stationary for 20 minutes), your false alerts go down but the bomb may have already gone off.
For faces, the challenge with 50-60 per minute is not the pure volume. You can specify sufficient hardware and set up analysis for sufficient frames per second.
The problem with a person a second is that it means lots of people are going to be obscured by people in front or next to them and that you might not get a clean face shot at all. This is even before you get into people wearing hats, or looking down, etc., that make this even worse.
For years, face recognition people have talked about 3D solving this problem, i.e., get a partial face and magically generate the full face but it still seems to be mostly a dream.
The problem you will find with face recognition in a crowd is that a handful of people will claim to do and you can get some results but once it goes to scale, the level of problems become so great that end users just resign themselves to it not being use the headache.
John, thank you.
Yes, I also thought that - if everything so good about iOmniscient why are they unknown to most of the world.
Is it possible that situation with analytics in crowded environments could change in near future? What do you think?
And do you know any Face recognition software that can accurately detect/recognize about 50-60 faces per minute?
I'm sorry for so many questions, but I'm new at IPVM and IVA has really big demand - here in Russia.
Those scenarios are science fiction. Recognizing faces or abandoned objects in crowded environments is nearly impossible (to do with any accuracy). There's just too many physical obstacles that make it an infeasible.
Iomniscient has been hyping analytics in a crowd for years yet every indicator shows that they are still a tiny company. If they could have come anywhere close to solving it, they'd be a massive company with huge revenues.
John, thank you very much for your answer.
In this case - what options (vendors) do we have if we need detecting abondoned object (not contrast) and face recognition at crowded scene (like metro/train station)? Is there any at all?
We have not tested iomniscient. Overall interest in analytics is weak. Plus, while iomniscient makes some astounding claims, we have yet to hear from anyone we trust that they work well. The best we can tell is that it's mostly marketing fluff.
Have you tested iomniscient at IPVM?
Unfortunately I don't have access to LinkedIn discussions about them.
Thanks John! I appreciate your feedback.
Tiago, iomniscient makes some very aggressive claims. However, as you noted, rarely see any real use of them. I spoke with them a few times including a half hour at the trade show where they had their face recognition in a crowd dog and pony show. Their answers lacked technical detail and included a lot of hand waving. I remain skeptical.
We had a few LinkedIn discussions about them - see 1, 2 and 3.
Here's my thoughts.
First, other than us (VideoIQ), I honestly wouldn't consider the companies you listed as "active" in the common analytics marketplace. Yes, they all have some form of product they sell as "analytics", but they are not making major waves (IMO).
In no particular order, I see good things from:
Bosch - They appear to take their analytics product seriously and put in major effort behind it.
AgentVi - They seem to have gone through some up and down cycles in the US, but overall they are also committed to their product and applications.
Sightlogix - Really pushing the thermal analytics angle. I think they do a good job of marketing to a niche within a niche (long-range thermal analytics).
Aimetis - I don't hear of them much in the US, but several of my international guys come across them.
PureTech - They are relatively recent, but again seem to be making a concerted push into the analytics space.
I don't want to color things too much with my personal opinions of any of the above in terms of price, performance, etc. But if you want to know the names that are heard more frequently in what I'd consider the top analytics deals/applications, those are the ones that have the most viable products, IMO. These are also the companies that appear to offer the best probability of delivering a usable system for a practical price. There are several semi-well-known and lesser-known companies that I consciously omitted because I would only recommend them to someone I was not trying to help ;)
This is also limited to "perimeter protection" focused products, which based on the initial question I took to be the main topic here. There are a number of companies that concentrate more on business analytics like Scopix, which tend more towards the power of their reporting and graphing functions than the overall ability to classify interesting/threatening and non-interesting objects optimally.
I wonder whether anyone could share some toughts about iOmniscient. They have a quite pushy marketing, but in terms of usage their video analytics products do not seem to be widely employed. Any experiences about them?
Well, ObjectVideo is more of an 'intellectual property' company than a production video analytics provider. I know they still offer their analytics for commercial sale but I have not heard anyone in a long time using their analytics.
Ioimage has really faded since the acquisition by DVTel 3 years ago. There's been minimal improvements to the product released and you no longer here about them often.
NICE continues to do its mostly large scale projects but beyond that they are not a factor in the rest of the market for analytics.
VideoIQ is big in the commercial space for video analytics with a lot of real world successes.
Others to consider include Sightlogix and AgentVi for pure play analytics (see a recent discussion on VideoIQ vs AgentVI vs Sightlogix). There's also Aimetis who builds them into their VMS.
Beyond that, there are lots of little analytics companies around the world but I'd be hard pressed to categorize most of them as 'worldwide top companies'.