Computer vision based security system for Multifamily properties

Multi-Family and Apartment Investing
8yFollow

Computer vision based security system for Multifamily properties

Krishna Chava
Specialist · Carrollton, TX · Member since 2016 · 145 posts · 100 votes

I am working with a partner in Silicon Valley to develop a computer vision based security and monitoring system for multi family properties. Please note, some of the items I propose below might look like invasion of tenant privacy which is a valid concern. I would like inputs from BP community to develop and refine the system such that multi-family property owners and property managers can get useful analytics. Please read on if interested and thanks for your time.

In brief, we will have cameras across the property (entrances, parking lot, corridors and common areas) that can recognize faces, and licence plates and use that information to develop analytics that can be used by property managers.

I own 70 doors and, based on my experience, I can think of following use cases. Please comment on them and and any other functions that would be useful.

Use cases:

1) Log car license plates going in and out of a complex with cameras at entrances. For each license plate in database, we will have complete history of times they entered and left every day. If any of the license plates are not registered with  the property, it can be investigated.

2) Log any people entering or leaving the apartment complex on foot. Since we can detect faces accurately , we will have similar database for people entering on foot.

3) Have cameras in parking lot and inform if someone or groups of people are loitering around the parking lot for more than 30 mins. We can analyze people movement and report incidents like this in real time.

4) Alert immediately if a police car shows up, hear police sirens in the property or hear gunshots.

5) (Controversial but likely legal): Monitor how many unique people are entering/leaving each door daily (from cameras in parking lots or corridors) and log that information. If there are 3 people on lease and system shows 5 people living in a unit, property manager will know.

Idea is to give property managers and remote owners analytics for keeping property safe and secure. It also helps tenants in case of break-in into their car in the parking lot and also could detect some illegal activities. The system itself will look like any other security camera system, therefore shouldn't raise any concern from tenants.

While the technology is developed and ready, business case makes sense at 1% of gross rents for monitoring and reporting. We plan to partner with local security companies for installation and maintenance of the system. 

Thanks again if you have read so far and I appreciate your inputs.

1Reply
51 views

Most Popular Reply

Account Closed
Specialist · Houston, TX · Member since 2016 · 68 posts · 41 votes
8y

I think you have a good idea in my opinion. I have two questions, one is a bit more technical and one more business related. 

I used to work at MIT on machine learning algorithms and decision support tools. The concept is a little different, but we essentially made tools that would have practical applications such as the tool you propose. 

From a technical perspective, one challenge we constantly faced was how to process large amounts of data with lots of noise. It sounds like the technology has been proven so this might not be an issue for you, but one thing we found was that our technology was much more difficult to apply in the real world rather than our simulated sample sets. A lot of this had to do with the pure volume of data - especially, when you are trying to predict or identify rare events. If you have not already tested this at a property, I would suggest that you quickly make an MVP and install it somewhere. Iron out the bugs early. I apologize if you have already done this, but I figured I would share my lessons learned in case it helped. 

From a business perspective, I think that 1% gross rents is pretty high for this type of product. This type of product would best apply to large properties in my opinion (100+ units), because they have the revenue to sustain these types of expenditures. Typically management companies charge 3-5% for their services. 1% for enhanced security seems pretty steep. 

1Reply
See this reply in the discussion

16 Replies

Jump to latestLatest
  • Account Closed
    Specialist · Houston, TX · Member since 2016 · 68 posts · 41 votes
    8y

    I think you have a good idea in my opinion. I have two questions, one is a bit more technical and one more business related. 

    I used to work at MIT on machine learning algorithms and decision support tools. The concept is a little different, but we essentially made tools that would have practical applications such as the tool you propose. 

    From a technical perspective, one challenge we constantly faced was how to process large amounts of data with lots of noise. It sounds like the technology has been proven so this might not be an issue for you, but one thing we found was that our technology was much more difficult to apply in the real world rather than our simulated sample sets. A lot of this had to do with the pure volume of data - especially, when you are trying to predict or identify rare events. If you have not already tested this at a property, I would suggest that you quickly make an MVP and install it somewhere. Iron out the bugs early. I apologize if you have already done this, but I figured I would share my lessons learned in case it helped. 

    From a business perspective, I think that 1% gross rents is pretty high for this type of product. This type of product would best apply to large properties in my opinion (100+ units), because they have the revenue to sustain these types of expenditures. Typically management companies charge 3-5% for their services. 1% for enhanced security seems pretty steep. 

    1Reply
  • Krishna Chava
    OP
    Specialist · Carrollton, TX · Member since 2016 · 145 posts · 100 votes
    8y

    Thanks for the reply @Account Closed. Founder of the company has already deployed the the technology in retail sector and most of the challenges you have mentioned can be overcome. He will create a BP account to reply directly. I am helping him identify a property for MVP, deploy it and collect data. Will keep updating the progress on BP. 

    You might be right that 1% is too high of number and actual price point depends on functionality achieved once we have a refined product.

    0Reply
  • Badri Varadarajan
    San Jose, CA · Member since 2018 · 4 posts · 0 votes
    8y

    @Account Closed mentioned, the technology itself has been proven in a different market -- we have existing deployments by international brands in retail who are using our technology mainly to measure shopper engagement with various displays and alert on out-of-stocks, etc. We have been exploring how the core technology may be applicable to property management and also to help manage convenience stores, gas stations, etc..


    Happy to take up a detailed offline conversation, but let me address some points here.

    1. Accuracy -- As you know, the accuracy of AI / deep learning based techniques has improved exponentially in just the last 3-4 years. Further, ours is a self-learning system where you can start deploying a few applications very accurately, and gain accuracy in other applications over time.

    2. Privacy -- All our analysis is local. We don't even need a broadband connection to deploy the product. Only anonymous metadata gets uploaded to the cloud.

    3. Cost -- As you can imagine, costs come down drastically with scale, but even to begin with, we can do MVP deployments at fairly low costs if we can identify use cases of interest.

    The exercise we are going through right now is figuring if there's a real pain point for property owners and managers. If repetitive monitoring can be done at lower cost (time or dollar), increased accuracy and without Big Brother storing all your data in the cloud, what would you use it for?

    Happy to chat offline, or get feedback from the community. Thanks!

    0Reply
  • Omar Khan
    Rental Property Investor · Dallas, TX · Member since 2017 · 1k+ posts · 1k+ votes
    8y

    @Krishna Chava Apart from massive privacy concerns (residential - even MF (only loan is considered commercial, RE is still considered residential) - is different than commercial) with a serious possibility of a lawsuit, who will pay for the data management/storage? 

    IMO, you will also run into practical issues when property managers will start asking prying questions that are unrelated to tenant behavior but which are based off the PMs biases (social, cultural, economic). This will also vary by state as each place is different especially around what data is public and what is private. 

    From a business perspective, I agree with @Account Closed that 1% of gross rents is very high. Unless, a Sponsor is investing in a war zone, you'll have a hard time convincing a B/C+ buyers about investing in your product unless there is more of an economic incentive. Partnering with local security companies might be a better route.

    The idea sounds great on paper, but I am interested in knowing what your pitch would be to PMs (especially, if this will reduce/mitigate the need to hire their services) and Sponsors.

    0Reply
  • Krishna Chava
    OP
    Specialist · Carrollton, TX · Member since 2016 · 145 posts · 100 votes
    8y

    @Omar Khan Thanks for your response. We can address privacy concerns better than a traditional security system that uploads data to the cloud. All our processing is done locally at the processing hub located in the property and only stats any analytics are uploaded to cloud. We don't therefore need a broadband connection. In the cases where snippets of videos is stored by the system (recording a break-in), this video will accessible to all tenants through a tenant portal. We plan to market this as adding security to tenants while not collecting videos data as any traditional security system.

    As for marketing pitch, we want to start with complexes that have 50-100 units where it's expensive to have reliable onsite staff. We don't want to market to war zone properties, that's not a good sell.  We want to learn from property managers and expand use cases look at stats like how many people are using on site pool, report any safety hazards at the pool, people using various amenities like outdoor grills, dog park, walking trails etc.,

    Having revenue as a percentage of rent might not be a good starter. However, we can make a value proposition of $10 per unit which is $500 per month, which is cheaper and more reliable than hiring a security guard. For offsite property managers, this enables them to reduce the number of trips needed to monitor the property.

    0Reply
  • Mark Allen
    Real Estate Investor/Broker · Irving, TX · Member since 2015 · 520 posts · 263 votes
    8y
    In areas where crime is eminent, I don't see how this stops the problem (crime). 99% of properties know exactly who the trouble tenants or know that petty crime will occur. So how do you limit crime? For example, a drug dealer will operate out of their unit. People will visit their unit and exchange money for drugs. The manager knows who the drug dealer is, but they can't just evict them for saying they're a drug dealer if said drug dealer is paying on time... or you can't put a camera in their unit that proves they're dealing drugs. If the owner pays the city police a nice sum of money, the police should stick around the property and not let loitering occur, especially to that drug dealers unit. That will in turn cut off that dealers business and he'll have to move. Am I making sense? That's just the main way I see putting a stop or reducing crime. Not technology, unless it's an armed robot :)
    0Reply
  • Omar Khan
    Rental Property Investor · Dallas, TX · Member since 2017 · 1k+ posts · 1k+ votes
    8y

    Interested in seeing local PMs response to this tool. I like the idea but find that most PMs are very resistant to change of any sort (this is a step above the usual).

    As an example - 100 unit building, $10/unit cost: monthly expenses = $1,000 --> annual expenses (or decrease in NOI) = $12,000 --> valuation difference on property (assuming 5% cap rate) = $240K.

    The average owner is making the above calculation every single time an expense (or decrease in NOI) is proposed.

    Again, I like the idea and think it has potential BUT the sales pitch/convincing stubborn people is I imagine having issues.

    0Reply
  • Rick M.
    Rental Property Investor · Fort Lauderdale, FL · Member since 2017 · 151 posts · 94 votes
    8y

    I've been looking at a security system for my complex, my concern is all the wiring and if someone was to cut one wire(s) that might compromise the system. But I also know that wireless signal can degrade causing poor picture resolution.

    I personally like all of the points especially numbers 2, 4 and 5.

    I can't see an issue with running security cameras in common areas, parking lots or corridors. I like the tenant portal idea. I'm very interested in this thread.

    0Reply
  • Michael DangPro Member
    Rental Property Investor · Houston, TX · Member since 2015 · 454 posts · 273 votes
    8y

    Interesting thread.  The data interpretation would be useful in a few ways I would imagine.

    For example, analyzing traffic and usage patterns for vehicles and common usage areas would be good to determine the anticipated lifetime of the asphalt parking lot and if sidewalks or walking paths are not being used and why.  If a unit has more than normal traffic then maybe budgets can be refined to account for units that will need more R&M then others.

    I wonder how can you leverage this technology to bring value add to the property by increasing Other Income?

    0Reply
  • Krishna Chava
    OP
    Specialist · Carrollton, TX · Member since 2016 · 145 posts · 100 votes
    8y

    Thanks everyone for the inputs so far

    @Mark Allen "The manager knows who the drug dealer is, but they can't just evict them for saying they're a drug dealer if said drug dealer is paying on time". I would extend this scenario to where property manager/security guard is either colluding with drug dealer or looking the other way. Either case, I am dependent on a part time employee with questionable loyalty. This is why 20-70 units apartments are not as popular as large apartment complexes with full time staff.  to me, having a computer vision system is like a trusted pair of eyes.

    Our quest is broader than a security system. Imagine a camera that is viewing parking lot; using our system, you can program several trigger events and change this programming based on usefulness of data. For example, you can look for people speeding in parking lot, people not picking up after their dogs, people throwing trash and also any loitering as I mentioned.  Similar can be said about camera facing a swimming pool.

    0Reply
  • Krishna Chava
    OP
    Specialist · Carrollton, TX · Member since 2016 · 145 posts · 100 votes
    8y

    @Omar Khan As any good investor, I wouldn't spend money on system that makes me loose money, not even a penny. This system should save money by keeping down property management costs (less trips to property), reduce security costs (don't have to pay for cloud backed up security systems that store large amounts of data) and generally making tenants feel more secure as small incidents are immediately attended to. 

    @Rick M. would love to chat with you about your specific security system needs. @Badri Varadarajan can comment more, but I believe cameras talk wireless to the local hub and picture quality would not be degraded as we are uploading data over wifi to remote server.

    @Michael Dang thanks for the input about using this for other income. This is where specific use cases are very valuable. I can think of a case where common area space is premium and apartment owners can analyze use patterns by experimenting with a grill, picnic area, dog park, child swings etc., and later analyzing use patterns.  

    0Reply
  • Mark Allen
    Real Estate Investor/Broker · Irving, TX · Member since 2015 · 520 posts · 263 votes
    8y
    Krishna Chava sounds like big brother is always watching... I'm sure turnover would increase if staff was scolding tenants for speeding or not picking up after their dog.
    0Reply
  • Badri Varadarajan
    San Jose, CA · Member since 2018 · 4 posts · 0 votes
    8y

    @Rick M., Wireless cameras should work fine as long as they are sufficiently close to the access point, or you've implemented a mesh system.  Please PM me and I can offer some advice. Today, just about every commercial establishment has a security system with cameras. Typically, they lie there recording video and only get accessed if there's an incident to report to insurance. More modern systems support (i) storage of video stream in the cloud with some minimal automated analysis, and/or (ii) remote access of a video livestream, but someone needs to sit and watch for activity.  Both of these require a broadband connection, and offer limited insight. 

    We are proposing that a local device will connect to the cameras or other sensors, analyze the data and only send up digests or clips.  There is a limit to what all can be analyzed, but modern techniques are pretty sophisticated in what they can analyze given sufficient density of sensors.  Obviously, we would start with one or two things which are of most value -- say for example detecting guns, cars & loitering; counting people entering and leaving some spaces; cars, etc. 

    As @Krishna Chava, this is cost-effective since we only need a cheap cellular connection and not a broadband one. 

    0Reply
  • Badri Varadarajan
    San Jose, CA · Member since 2018 · 4 posts · 0 votes
    8y

    @Mark Allen, @Omar Khan,

    You've both rightly brought up the issue of privacy. The last thing tenants want is to be under the magnifying glass in their own homes or neighborhoods the way they (we all) are in any public place or with any activity online. That's why we believe local processing and completely open systems are a win-win. Any rental home sensor system can, should, and in our case, do support identity-based access. For example, a tenant has access to the same information that the property manager or owner sees from their section of the property, but has no access to data from anyone else's section. 

    A well-run and secure property is in everybody's interest. We are only trying to see if technology can bring meaningful assistance to this problem. Obviously, an obnoxious, nosy property manager will drive tenants away, but they'll do so anyway, with or without hard data. 

    0Reply
  • Badri Varadarajan
    San Jose, CA · Member since 2018 · 4 posts · 0 votes
    8y

    @Michael Dang, great point. Indeed, the other income is our primary value prop in our first line of business -- which is business intelligence for brands in retail. Would love to hear more about adjacent value that property analytics can bring. 

    0Reply
  • Mark Allen
    Real Estate Investor/Broker · Irving, TX · Member since 2015 · 520 posts · 263 votes
    8y
    Seems better fit for retail, IMO.
    0Reply
Join the conversationCreate a free account to reply, vote on answers and follow this thread.
Log InJoin for Free