Business Continuity Series: Video Analytics Solutions using Meraki Smart Cameras


Learn from WWT's Kait Miller in this 19-minute video about the many use cases for video analytics solution built using Meraki cloud-managed smart cameras and integrated with analytics software.

Please view Transcript below:


Kait Miller:  Hello and welcome everybody. Today we are going to talk about Meraki MV smart camera solutions and video analytics that we can overlay on top of a Meraki MV solution. My name is Kait Miller and I am a business development manager with World Wide Technology, with a focus on Meraki Smart Cameras and video analytics. So let's talk a little bit about Meraki MV smart camera features just at a high level. Native capabilities within the Meraki MV cameras allow you to create video walls right in the Meraki dashboard. And you can have up to 16 streams per video wall with multiple video walls able to run. There are flexible storage options. So you can store the video local to the camera. You can back that video up to the cloud, or if needed, you can store that video footage on-prem as well. We have the ability to send motion alerts so we can create digital rails for areas of interest with this as well.

  So if there is an area within the field of the view of camera, that should not have anybody in there, we can set up a motion alert in that area and alert if somebody does enter that area. The other thing that we can do with motion alerts is set up schedules. So if a business closes at 6:00 PM and there should be nobody there after hours, we can set up a [inaudible] that will send alerts if there is motion in that business, after those hours of business have closed. All cameras have the ability to be run on wireless for quick and easy deployment. The other option is to utilize any standards based switch that has POE capabilities and run an ethernet cable to that camera, to power it and run it.

 And what powers all of the awesome analytics use cases that we are going to talk about is the Meraki MV Sense API. So the MV Sense API has people detection, vehicle detection, and lighting level detection. You can make a couple of different calls to the API, either for historical aggregate information, so how many were here at X time in the past, and also for current snapshot information, so how many were here now. And this is what we call a snapshot API, and this is the majority... Everything that we're going to talk about today, the snapshot API is really what's being used for the majority of these use cases. So let's get into it.

 WWT and Meraki have partnered with an ecosystem partner called EveryAngle and EveryAngle lives in the Amazon Web Services cloud. And what EveryAngle does is they consume the Meraki MV Sense API data. They bring that into their cloud instance and they then utilize their own machine learning capabilities to determine business outcomes for customers. Along with their own machine learning capabilities, they also have the ability to utilize AWS recognition or other machine learning capabilities that live in the cloud as well. And this allows them to determine things like a person's age, or gender, or emotion as they're engaging with specific areas and with specific businesses. And so what I want to do now, is just go ahead and go through some of those very specific use cases, how our customers are using them, and how they can be applied. And just start to get you thinking about the art of the possible.

   So one of the first capabilities with vehicle detection is the ability to do license plate recognition. So as you can see in this instance here, this is my truck. This is my license plate, and I set this up just on my driveway, went ahead and activated license plate recognition. And every time I pulled in or out of my driveway, there was a detection and it read my license plate and populated a dashboard with that license plate. So this is not new technology, license plate recognition has been around for a while, but I do think with some of the recent changes to how businesses are operating, there is a new application for this that World Wide Technology can help you with.

  And what comes to mind for me is enhancing the customer experience for curbside delivery and pickup. So every instance that I personally have engaged in for curbside delivery, and also all of the instances that I went out and looked up online for curbside delivery, include a manual effort by that customer when they arrive on site. They either need to call the number on a sign and then wait, or they need to engage with that application and let them know that they have arrived after parking. Let them know what parking spot they've parked in, and then remember not to go and move their car. And where I think we can really have a huge impact to truly enhance that customer experience is by utilizing license plate recognition and creating an alert to the store or location employee that looks something like this.

New vehicle detected. Here's a picture of that vehicle because it's all well and good to say, "Hey, it's a gray SUV." But if that person's not a car person, they may not know exactly what they're looking for, or exactly what the shape of that vehicle is when they get outside. So this gives them a visual representation of exactly what vehicle they're looking for.

    It also gives them the license plate so that they can very, very strictly confirm that this is the exact right vehicle. Products that have been prepaid for are not going to the wrong person or to the wrong vehicle. Can also identify the color, make, and approximate year. We can identify which camera is sending that alert as well. So the employee will have a general idea of where that vehicle is parked, again, without customer interaction. This is all just going to be automated for them to enhance that experience. And we can call out a zone as well. In some cases, the zone may be a specific parking spot number or parking spot label. But a lot of times this is where we see things start to separate from an idea of how we expect something to happen versus how humans actually behave.

 And so one instance of this, that actually a colleague was telling me about, he went out to a store to pick up some goods. He pulled into the parking spot, pulled out the app said, "I'm in this red vehicle. This is the parking spot number, I'm in seven." And he then waited, and the wait began to be 10 minutes and then 15 minutes. And after a while, he started trying to twist his neck to see inside that location. And then he decided he was going to move to a different parking spot to get a better view of what was going on inside of that location, to see what was taking so long to bring his delivery out to his car.

                                    So then after some time had passed and the employee was finally bringing his order out to his car, they walked over to parking spot seven. Obviously he was no longer there, and there was some confusion, some flagging down that had to occur. So I think this is just an area where we can take this entire manual human error laden process and bring some automation to the table to really enhance that customer experience and give a better end to end experience overall for both customers and for employees.

                                    Another use case generally for analytics is to utilize analytics for safety. And when we start to talk about utilizing analytics for safety, we're looking at potentially warehouses, or manufacturing, or construction type of positions in which we want to ensure that an employee is wearing the proper work equipment before entering a workspace. So if somebody is supposed to be wearing a hardhat or reflective vest and boots before entering a manufacturer's warehouse floor, we can utilize computer vision analytics to go ahead and integrate that into access control systems. And with that, when that employee goes to badge in, they can actually be denied entrance from that facility if they are missing a piece of this safety gear that they're required to wear. So when identifying personal protective equipment, I know the first question that's going to come up is, "Hey, can we accurately detect medical PPE equipment for the current environment that we're all living in, where everybody needs to wear a face mask for the safety of all?"

                                    And the answer to that is, this is very possible. Cisco Enterprise and EveryAngle are working together to develop an entirely new ML that will accurately detect the PPE face masks across all environments. And the reason that they are updating this from what exists already is because we see that a lot of people are straying away from typical N95 or hospital type PPE masks. And they're using scarves, or they're pulling up a turtleneck over their face, or they're using cloth masks with different types of designs on them. And so there are a lot of images to feed and allow this machine learning to accurately detect these. But again, we just want to make sure that that's updated and able to accurately detect all different types of PPE masks across all environments.

                                    All right. So now let's talk about in essential retail or as retailers are returning to the workplace and trying to reopen, some of the COVID-19 use cases. And they really boil down to two things, we're looking at employee safety and customer confidence. And that's really what it comes down to when it comes to COVID-19 use cases. You want to ensure the employees that you are bringing back into the workplace are going to be safe. And you also want the customers to know that you are taking appropriate steps to keep them safe, so that they have the confidence to reengage with your locations. However, this technology is not only for COVID-19. We have plenty of non COVID-19 use cases so that this technology can continue to be relevant and continue to be a good investment as we move into the new normal.

                                    And some of those are accurate people counting, identifying wait times in line, looking at dwell times in front of a specific self service kiosk, or a specific display, identifying customer emotion. So both when they enter a location, leave a location, or while they are engaging with a self service kiosk or another display that's set up in that store. And also for security. Now, these all boil down to longterm outcomes and longterm goals. One could be to take a self service model and be able to deploy that model nationally, but we can work together and determine what the best use case is for you and for your organization.

                                    So how do we accomplish some of these things? So for some of the COVID-19 use cases, we can utilize an application called Physical Distance Controls. And what this does is we put a camera facing the entrance, and we also put a camera at the exit facing inside of the store. And this allows us to count people as they enter the store and subtract people as they exit the store. This will give us a total store occupancy number. From there, we want to identify a safe occupancy for that location. So let's say the normal safe occupancy is 100 people, and the CDC is recommending 25% capacity for all locations in a specific state or town. We can set the safe occupancy of that location to 25 people, and we can then measure against that statistic.

                                    One of the ways that we can show this information is utilizing digital signage. We're starting to see that digital signage can become very useful in these locations where we're utilizing this application. So we can just show a green thumbs up that says "This store is below its safe occupancy number you can enter." or we can show a red stop. "This store has reached its maximum safe occupancy. Please wait before entering." We can also integrate this technology with some speakers. So if you're the kind of person who has your head down on your phone, like I do, as you're walking around texting and looking at different applications on your phone, a speaker system can audibly alert somebody to recognize the digital signage.

                                    And when we go back and look at the video that's occurring where these systems are in place, we do see that these speaker systems tend to have a huge impact. Because not only does it alert that customer to recognize the digital signage before they enter, but they then start take note of any other visual cues that are in that location. Such as tape on the floor indicating one way traffic down specific aisles, or how to interact with that location for safety.

                                    So, I would encourage you to go to wwt.com, it's our platform, to find more information about the Physical Distance Controls application. This here is an article that I published, and if you go into wwt.com, hit the search bar, and search for "Business open faster", you will find this article and a bunch of other content related to how WWT can help you open safer, sooner.

                                    So how do we ensure that this investment is good moving forward? Next Gen Footfall is another application that we can run on the exact same cameras that are already in place for the Physical Distance Controls app. And the Next Gen Footfall is really just a fancy way of saying highly accurate people counting. We can look at conversion rates and integrate that into a POS system. So you can start to determine which demographics are purchasing which items in your location. We can look at dwell times, so how long is somebody spending in front of a specific layout or design, and how useful can that information be to marketing on whether or not you should run the same promo again in the future. And we can measure bounce rates, maybe very often somebody is entering and exiting the store very quickly. How can you change the visual display at the front of that store to stop people from leaving so quickly and encourage them to further engage with your location, to the point that they then make a purchase.

                                    Suspicious person detection. So this application is going to detect for anybody covering their face. And I know where we're living today, everybody's covering their face with face masks. So again, that ML is being updated to detect whether that face mask is of suspicious or nefarious activity or due to the health risks, and the typical face mask that we're seeing everybody wear today. This application also has the ability to detect weapons. And there are a series of different types of masks and weapons that can be toggled on or off. Because if you walk into a hardware store with a hammer, it's probably not very suspicious, but if you walk into a bank with a hammer, it's a lot more suspicious. The other thing that we are able to do with this specific application is to set specific percentages at which you want to be notified.

                                    So if you have a location that is very high risk, and there's a lot of issues at that location, we can set that to be "I'm 30% sure this is suspicious activity." So we're going to notify you because you would rather work with a few false positives than miss something. And if you have locations there's never any issues at then we can say, you only want to be notified if we're 90% sure that this is a suspicious person that needs to be checked out. So very customizable and flexible for all of these applications.

                                    All right. In closing, I would like to know what is your impossible ask? I encourage you to visit us the digital workspace portion of wwt.com, and thank you very much for attending. Have a great day.