Startup Memoreyes Wants Feedback

Hi John - Loading to YouTube. I thought Google drive would work. Where do I send the link?

I'll change the name... Just trying to stand out from the crowd a bit. - Best

Hi #1 - Yeah, it's boring, and it's a long story. Understood. I did this on IPVM specifically to hear peers reactions before I spend the $$ to let a "professional" tell the story. I wanted to know your questions/perceptions first. I'm too close to it. Then we can craft a message.

Yes, we're going to align with a video central for response.

And... this isn't nearly as expensive as one might think. More later.

Thanks for taking the time. Best,

also, in your video from 5:13 to 5:30 your name is spelled wrong in the title at the bottom left

Thanks Thomas - the video was recorded by a Dahua NVR. We are currently using a Uniview starlight PTZ. Both are ONVIF compatible. We set h264 with extra Uniview compression. We do 15fps, 2MP and average a 2-3mbps stream over the wireless links.

This will work with any good ONVIF compatible PTZ. We're about to begin tests with an AXIS.

We have our own "NVR" on the local network primarily for recording the local camera director actions and coordinating remote camera control from the cell tracking system. All VPN connected.

We have a few dozen PTZ record with well over 3 million "turns" over the past two years. We're figuring about 5 million turns MTBF on the PTZ steppers and drives give out. At night there are few turns as there is little activity. During the day a camera can get very active.

Thanks Dwayne. I am believer in the old Marine saying of "No guts, no glory". And I completely agree that at this moment the only pertinent (mostly) comments come from the industry folks here.

I also fully agree that we need a better presentation. But I wanted to get raw results here before planning / spending for it. $10k - $15k is still a lot of money for a small company, but we are already scripting a more professional presentation.

It is ONVIF compatible. We just look like another camera to the VMS.

As for our location/sensor data we are holding that on our own local computer. There's no import formats for the data we collect. Genetec has been kind enough to begin those discussions. I haven't even approached Milestone.

There will be lots of polishing to do but the foundations have been laid.

Best

Also, sorry, please contact me off line about a trial. Bottom line we'd be thrilled. We've done a lot of our own tire kicking and it's high time others have their chance.

jcollings@memoreyes.com

Best

Somewhat..

I won't comment on others work.

I will say this, ask them if their system detects a human's motion 360 degrees, and then turns and zooms to frame the target. Also, ask them how they deal with hills, a gun shot, other noises. Do they have a speaker in the system so a person can talk to the target in question?

We appreciate the interest.

Best,

Hi #4 - It means I've developed a number of products and services before and.. I don't ring the dinner bell before dinner is ready. Our team doesn't do vaporware.

There are many developments in this system. It's taken over two years to perfect them, or get them ready life/safety commercial release.

If you want to talk more about the systems preparedness I'd be happy to. My email is above.

Best,

Hi - thanks for taking the time. Great questions.

"What problem are you attempting to solve with your product?"

At a very high level we think one of the jobs of surveillance is it should deter reckless or criminal behavior and failing that should produce video documentation that provides authorities legally actionable evidence to address events. In wide area outdoor environments, we think this can be done much better than current industry methods. Specifically, the cost of deploying enough quality cameras to effectively watch over acres worth of territory, given current methods, is prohibitive. Thus, the industry, understandably and rightly, focuses on entrances and interiors. - We think we're addressing the problems of outdoor safety, at an installed price almost any commercial/municipal property can afford.

We think one of the reasons RING is so successful is when a potential criminal comes to the door and see the device they know they’re being videoed. The yard sign effect. We think most current outdoor surveillance systems do not evoke deterrence. People don’t know they’re working, or don’t notice them at all. Thus, the industry still gets many videos of reckless and criminal behavior.

Additionally, we believe that most of the video evidence gathered by outdoor surveillance cameras is marginal in its investigative and evidentiary effectiveness. Either due to simply not recording the event, or if it did, poor quality, primarily due to distance from the camera.

So, we believe we’re solving, reducing both these problems.

Deterrence

• Our physical presence outside is different. Looks different, acts different. The device/system has a physical presence outside that, because people quickly realize the device notices at them autonomously, at great distance, deters bad behavior. On our test properties crime and petty damage has dropped 20-50% year on year, some more.
• We’re told the device doesn’t offend residents/employees/customers because it doesn’t look “institutional”. We’re told it looks “intelligent”, evokes a feeling of safety. Thus, we’re providing the effect of a socially acceptable onsite guard for a fraction of the cost. Did I mention it comes in 5 colors? 
•  
• Detection and Documentation
•  
• Ever had a PTZ on patrol see the suspect, the event, only to look away at the worst moments? When you direct the camera, make it zoom in, dwell on specific motion or noise, you/we tend to see the event, or at least get a good picture of the suspect.
• With the addition of cell phone tracking a person can immediately call the attention of the system. Detection, documentation and whole new level of deterrence. 
•  
• Cost
•  
• $6,200. What’s a guard cost?
•  
• Thoughts?
• Thanks,

Hi - please see my initial response to your good questions below. I forgot to answer #2 and #3. I got caught up in the psychology and technology..

" Are you concerned your product could be prohibited in states/countries where consent to record audio is required?"

In the 2 party states we do not record the audio as part of the primary stream. In a number of evidentiary examinations of our systems evidence we have successfully presented to courts that the audio direction finder was not "listening to the private conversations". Rather, it was reacting to noises of certain qualities for the purpose of directing the video camera. It was in no way interpreting or recording the verbal, private, conversations.

On the other hand in regard to GDPR, … our system definitely has intention. It's not incidental, as are static camera/audio systems. I can see the possibility where in certain countries... we're not allowed.

" Tell us about a recommended topology for deployment."

Higher density living/employment/retail, larger area (2-6 acres per camera), ground level (we don't look up except due to terrain/elevation changes), open versus lots of foliage, parks and parking lots. Apartments (our test environments), business parks, retail parking, corporate campus (outside), municipalities common area, parking lots.

Best,

Yes, the clients knew, know it is a work in process. Yes, we used zip ties to secure the radios at times, stainless straps now. Sluggish... yes, there's a delay, but the reason we're still on all those properties is because it's worked well to very well for them. Read my response to End User #5 please.

The cost? We're renting it to those apartment communities. $120 per month per camera. Average property has 7 cameras.

Which property did you see?

We started over 3 years ago testing individual components so things have come quite a ways since then.

I'd be happy to show you the latest.

Thank you - I look forward to talking. Best,

Hi - it's "starlight" as long as it's in color. .005 lux at f1.6 I believe that's generally in the range of the definition. IR after that.

No, I don't believe we've infringed on anyone's patents. If you do please send them/post here.

Yes, there's more work to be done. Agreed.

Best.

I wondered when someone would ask this. Excellent question, and goes to a core reason we developed our “system”. We don’t see this as a camera, but a system.

The short answer is area served. Our system serves a much larger area by being able to put more pixels on a target and a greater distance, thus one “camera” can do the job of dozens. A guard with binoculars versus a guard without binoculars.

First, let me say, I do not mean to disparage any manufacturers products, not Arecont, or Axis, or Avigilon or anyone. They all do quality work as far as I’m concerned. A multi-sensor device has many good and valuable uses such as a high traffic areas. But it’s “pixel on target” range is far less than the system we designed. It just wasn’t designed to do the same job. Also remember, these devices (great majority) have no built in IR illumination when needed.

As you suggested, let’s use the Arecont AV20275DN-08 for the example. Four 5 MP sensors with 8mm lenses. Those lenses have about a 40° horizontal FOV, for a 160° total. First, our system covers about 330° (we can’t look back through our pole), so the Arecont has almost half the “area viewed”, BUT it does this ALL the time, and has 4 times the bandwidth and 4 times the storage.

This is going to come down to the application. We’re going to say that our system serves a whole new category of application, and it’s just not proper to compare the two.

To continue the example, let’s say the goal is to provide 50 pixels per foot of resolution on targets. Not facial recognition quality for sure, but a good, useful picture.

For a 2MP sensor that’s ABOUT 20’ horizontal, for a 5MP sensor that ABOUT 25’.

Doing the isosceles triangle math (FOV) for 50 pixels per foot resolution with a 5MP sensor with ~40° FOV you get an effective distance of ABOUT 40’ from the camera. Sure, you get video of events further than that but its investigative/evidentiary value drops substantially as the distance away grows.

Our system uses a 22X optical zoom PTZ that can readily achieve 50 pixel per foot resolution at 300’.

40’ radius (πR²) X (160/360) =~2,500 sf. 300’R area X (330/360) =~ 250,000 sf. About 100 times more area, 100 times more cameras, installations, bandwidth and storage.

Everyone who’s still reading is screaming BUT the multi-camera system is doing this ALL the time, missing NOTHING. Very true. But nobody (few) would ever buy/operate the multi-camera system in this example.

It comes down to can a PTZ camera(s) be made to pay attention to, and video capture, the vast majority of pertinent human events over a very wide area. Our answer is, in many useful applications, yes.

This is where the newly developed sensors coupled with AI comes in. That’s what we think we’ve done.

If you’re still awake and willing, please read my response to Ross below. Thanks

Great, pertinent points and concerns. I’ll address all below but the answer to what happens between 7:30 and 8:30 is … the system can be very busy. You should see it when school lets out.

We don’t purport this to be perfect, just a major price/performance advancement in covering large outdoor areas, 5 to 50 acres worth. Please see my response to Abaas above.

Also, the cell app use is a work in process at this time. As experience teaches us, we may alter the way it works. For today it is designed to call specific, varying levels, of attention to yourself.

Liability – Anyone who chooses to download/use our cell app will agree to our EULA (End User License Agreement). All apps have it. This is very similar to a standard alarm company agreement. Best efforts, standard fitness for use disclaimers. As for deterrence value the app user can make the close cameras chirp, telling others they are being watched. When we implement/partner with a video central(s) we will evoke live 2-way response.

The following are long answers. I’m providing highlights. The system is designed to have more than one camera. Each camera always has another within view, typically 120-200 yards away.

Concurrent Multiple Cell App Users – The app currently has 3 levels of use, notice, “nervous” and duress. 30 second glances to continuous. We allocate 1 duress a month, after that we’ll ask for a “false alarm” charge. We know who ever app user is. We allocate 4-5 seconds per target. Some targets are offloaded to adjacent cameras when possible. At moments we’ll choose to look at “groups”, not individuals. Simply pull back the zoom. Something is better than nothing. We haven’t implemented it but there is a design for “system unavailable”.

Busy Scene/Multiple Targets – This, and question 5, gets into the AI, and the quality of the sensor data. This is where we succeed, or fail. Also, this is will be an evolving area. The system is designed to be automatically updated as get “smarter”. At a high level when the system has multiple targets we try and do what a human does, plug back our gaze and look around when needed. Most areas have 2 camera coverage. Look at a specific target when it becomes unusual, a scream, a running target.

Please remember, this is for wide areas, and at a cost that’s not hundreds of cameras for end users.

Also remember, most crimes don’t occur in front of crowds, and when they do everyone’s gathered around and screaming. That’ll get our attention.

Weather – We are not using pixel analysis, radar, acoustic and cell GPS. Right now we seem pretty weather agnostic except for audio. Rain can make acoustic data useless.

The $120/month is in lieu of purchase.

The $6,200 does not include installation. We charge ~$800 per device, that includes concrete pad, power supply (low voltage) trenching, and commissioning.

The $6,200 does include the wireless link equipment, pole and power supply. Obviously includes the PTZ and "camera director".

Currently, the system does not proactively create alerts, but it is intended to. More on that later.

While I may have responded a bit aggressively to some, I am deeply appreciative of the vast majority of these questions and comments. I've found them very helpful.

Hi Max – I looked at your offering. Very nice. Certainly, we’ve thought about the same problems. I like the BT beacons. We use them, but for a different reason. Let’s talk offline. jcollings@memoreyes.com

But, at a high level:

• Economics – in a pure rental/service we plan on replacing/servicing the camera about every five years. Ensure the customer has the best tech. If we/dealer shows up to replace one, we’ll replace them all. We expect about 5M turns. We have a built in daily positioning calibration routine and can 1- ensure it’s looking where we want it to and 2- gauge wear. We’re OK with the economics.
• AI… All of the above. That's the subject of a number of claims.
• Comms. Yes, wifi band. We mostly use the DFS channels. We ping regularly. Even with multiple hops we're typically at sub 5ms. If someone goes away we kick them locally/remotely. Given where we are we mostly have clear channels. There are alternatives. I'm excited about 5G.
• Edge/central… both. We have power. We don't deal with the issues you deal with.
• I like your feature. Not sure if we can object detect a gun reliably. 😉

Hi Gary - the radar is radar, RF. But it's FCC part 15, substantially less emitted power than a cell phone so, hopefully people won't be worried. The wifi could be jammed, any radio can be but I/we feel there is low odds of this. We're not trying for DoD applications, quite the opposite. The camera director (one with every PTZ) has 2 computers onboard locally, so if the market demands local storage we can offer it. Or, if the customer wants they can cat 5/fiber hardwire the cameras. Thanks,

Hi Lynn, interesting questions, thanks.

If I’m understanding your questions correctly you might be looking at us as standard pixel analysis camera direction. We’re not that. Very different.

1 – We could prioritize sensors. We never really thought about allowing users to do this though. For us the goal was set up the sensors/AI so.. it just worked, paid attention to the right thing(s) at the right moment. I’d be interested to hear how you’d prioritize events, and your goal for prioritization. The current config is set for “remote” user prioritization, track a cell user, or a remote guard using our camera. After that it depends on the combination of local radar, acoustic target qualities and the recent history of views.

2 – More is possible. It all depends on the current target load on the system but if an app user indicates duress they get priority. If more than one at a moment do we evoke partner cameras and after that load level as best we can on a 5 second cadence plus local audio, who’s screaming/shooting. Please remember, we know who did this as we know the cell number so if someone’s “gaming” the system they are part of the investigation. Additionally, it depends on how many cameras have low obscuration views of targets at any one moment. It’s… complex. So is life. Better than nothing, or low resolution nothing.

3 – The answer to this question is yes, but even deeper than you suggest. We keep/have a whole new level of search criteria meta data for the video. It’s a matter of tagging that data to the specific video segment. We will do that with our own NVR and desire to link that to others. We’re just trying to get better documentation of pertinent events. Document the truth.

4 – We have thermal capabilities already. We have a really neat 360° thermal sensor we developed. We were using that until radar became available which we believe is better overall. We’ve a design for using both but not at this time. Thermal can focus in on still human targets, good quality. But please realize, that on a spring/summer/fall day looking over roads or parking lots, a human is almost impossible to pick out at 50-100 meters, without really expensive sensors. We’re trying to make this system affordable/effective for any commercial customer, and parking lots are real important.

5 – Currently we have no interface for other sensors which would tell us “look here”. Suggest a use case please.

6 – Weather. Please remember, this is not video pixel analysis guiding this system. Rain does reduce/eliminate the audio direction finding/analysis side. I’m not sure I understand your term of “video noise” unless it’s in reference to issues with motion detection with standard video analytics during rain/snow. If that’s what you’re referring to we have virtually none of that.

This is a different way of doing outdoor safety, surveillance. Not perfect, but substantially better/less expensive than the status que. One side is passive, make the PTZ(s) autonomously pay attention to the right thing at the right moment(s). The other is active, an app user saying watch me, help me. All at an installed cost property owners can afford. Better safety for property and persons. Bar that for property and persons, documenting the truth while preserving privacy. That's the goal.