Beesource Beekeeping Forums banner

1 - 15 of 15 Posts

·
Registered
Joined
·
10 Posts
Discussion Starter · #1 ·
Hi,

Hoping to create a conversation around technology and beekeeping. I grew up with bees, am a hobbyist beekeeper and work in a medical robotics company by day. For the last year my team and I have been working on a tech to bring healthcare analytics to bee hives.

Hobbyist and commercial beekeepers in our community were saying they'd have generally healthy hives, and would often find some dead at the next inspection - we wanted to create a video record of what was going on, and make that useful by turning the video to data on a simple graph, so you only had to look at the videos of interest.

Our approach uses a camera and computer to visually monitor what goes on at the bee hive entrance all day. Activity is measured by tracking the number of bees flying in front using onboard computerized image processing. Video and data is viewable remotely through an analytics platform, where other data and hives are available to see too.

Beekeepers in our 15 hive beta program used EyesOnHives to monitor and take action to manage colonies suffering from queen failure, ants attacks (yes, that's a thing) and robbing. The ants were a real surprise for us - they really can and do take down even a strong hive when they want to. Having the early warning through video has enabled us to provide a safety net for other beeks wanting to see videos of other hives, and finding things like ants attacking.

We wanted to hear what the community had to say about our approach to non-invasively monitor hive activity. As far as we know we're the first to achieve this at scale, and have seen some pretty interesting behavior patterns.

Please let me know if you have any interest in hearing more, or questions.

Thank you!

Kelton
 

·
Registered
Joined
·
1,310 Posts
As far as we know we're the first to achieve this at scale, and have seen some pretty interesting behavior patterns.
Can you give some details about the interesting behaviors patterns observed? Thank you in advance for your attention and success for your project.
 

·
Registered
Joined
·
10 Posts
Discussion Starter · #3 ·
Hi Eduardo,

Thanks for your interest - here's a couple of pictures showing the behavior patterns

One of the most exciting things we found was what we call the "orientation spike" in data.
Orientation.jpg
Healthy Orientation Signature

In a healthy hive we see an activity spike mid-afternoon as new field bees take their first 'orientation' flights.

Sick hives, or hives that are under attack, show different activity signatures.

Robbing.jpg
Robbing Signature - when other bees attack a weaker hive.

Graphing the average daily bee activity over time shows how a hive is growing.

Trend Up.jpg
Growing Hive Trend

It turns out the flight activity is an indication of hive health as it grows and gives an early warning if something is wrong.

Queenless.jpg
Queen Loss and Colony Recovery
 

·
Registered
Joined
·
288 Posts
I would be very interested in this solution. Have you thought about incorporating internal hive temps and moisture? I have a lot of questionS. PM me.
 

·
Registered
Joined
·
5,540 Posts
It would be interesting to see how this works in much-colder climates than SoCal. It would be the same in warm weather, of course, but as the flying season changes (ramps up and down in Spring and Fall) the positive and negative trends would be different, though still entirely normal for the area. The intriguing thing to think about is aside from the obvious gross seasonal trends, what sub-variations might be observed.

I'm thinking of something like an apparent drop in flying in the spring (below the immediately previous level) that could also indicate a ramp-up of brood that needed care and warmth inside the hive vs a week or so earlier when the demands were less and the urgency to fly again was higher. In this case the downward trend would be an indication of good things to come, not the harbinger of trouble.

Very interesting concept. In addition to internal metrics (temp/ humidity) I'd add an external data set, too. What the heck, how about something that also listens to the hive's acoustic state, (pitch, volume, etc.)?

What would be very useful would be the discovery of a time of day (or external air temp, or some other factor) when a beekeeper could observe/monitor the hive for a short period to get an idea what might be going on with the colony, and why.

I look forward to hearing more.

Enjambres
 

·
Registered
Joined
·
10 Posts
Discussion Starter · #6 ·
I would be very interested in this solution. Have you thought about incorporating internal hive temps and moisture? I have a lot of questionS. PM me.
Thanks for your input John! I have PM'd you. Regarding internal temps and moisture - we wanted to keep this version of the system as simple as possible, and ideally outside the hive. In our experiments we found that anything inside the hive would eventually get covered in propolis!
 

·
Registered
Joined
·
10 Posts
Discussion Starter · #7 ·
Thank you Enjambres,

You bring up some very interesting points! Absolutely agree that the climate and location will play a huge part in what's normal for a hive - would be fantastic to get a profile for different areas that beekeepers could locally compare. We have geolocation of the device in the analytics platform, and pull in data from local weather stations to overlay on the basic activity data.

We'd be very excited to see what trends are present with regard to brood levels and drops in flight activity. One of the key ideas is that even if it's not known what's necessarily normal, if there are significant variations between hives, and many hives are monitored, it becomes easier to spot the outliers.

Pretty interested that people are talking about the internal metrics, and also audio. We have been testing all three of those but finding there are some implementation challenges and variables that are harder to control between hives. E.g. if you have a brood temperature sensor for one hive, and are comparing that data as the brood nest moves over time, you might see changes that are quite normal, but lead people to unnecessarily opening the hive and disturbing the brood nest. Not to mention the propolis. You are motivating us to take another look at this - might be something we'll bring into a future model!

The graphs that you saw above actually have videos behind them, so if there's a behavior of interest, you can see what's behind it. I just uploaded a demo of this to youtube here: http://bit.ly/EOHSmartphone
 

·
Registered
Joined
·
61 Posts
Kelton, hi. I've been an Open Source software developer for 20 years, and helped build the first streaming video over internet service, among other things. It took a long time, but I'm really glad it got to this point where you were able to do this Kickstarter!

Since video is uploaded and analytics are done on your server, I understand that is your value and secret sauce. And $10/month sounds very fair. Since your value proposition is mainly in the analytics, is there any chance of providing my own camera and streaming video to your service for analysis? I know 14 days sounds like plenty, but I'd like to keep 7 year video records for each hive.

Also, will there be an API so I can download analytics data and do my own display with it?

Also, is that $10/hive? As I bump it up from 2 hives to 20, that could get really spendy. Will there be volume discounts?

Also, thefts are getting more frequent in my area.

http://www.cbc.ca/news/canada/british-columbia/hives-containing-500-000-bees-stolen-in-b-c-1.1290800

Any chance of including in the analytics "hey, the hive is gone!" and also some camouflage for the camera itself? That would also cover the case where high winds topple the hive.
 

·
Registered
Joined
·
10 Posts
Discussion Starter · #11 ·
Hi Mycroft,

Thank you for your encouragement, and it's great to hear you're a fellow developer as well as beekeeper! Thank you also for contributing to the open source community.

These are some great questions!

First off, EyesOnHives is actually doing analytics on the device as well as at the server. EyesOnHives runs algorithms on the devices themselves to quantify the bee activity and determine whether the video should be uploaded in addition to the data. This minimizes bandwidth, so minimizes cost to both the beekeeper, and the cloud system. We found early on that most people don't have the signal strength in their backyard WiFi to support constant video streaming. On the server side we do further analytics, classifying activity over the day, and correlating this to other hives and over longer timescales for other insights.

We have explored some concepts around enabling makers and hackers to join the platform and use their own devices, it is something our team is interested in, but at this stage we do not publish our API or support this. We're actively discussing a project with a University of California Ag/Computer Science research group which may change some parts of this in 2016. We also have a non-released system which allows for personal cloud storage of years of video. I'm curious, how often you are looking back over your own archives?

For pricing of more than one hive, we will have some updates in the new year. We are working through the best way to align our goal of helping beekeepers monitor as many hives as possible, and giving beekeepers the best alerts and playback capabilities! Our crowdfunding to join the Beta group just closed, but this is one of our main goals in getting feedback from more users of the system!

Sorry to hear about those thefts, it's a heartbreaking thing to lose a hive, let alone lose a whole apiary to theft from a beek thief. Our analytics would actually show this pretty quickly to have occurred, and in addition to the 'per hive' monitoring system, we have yet another research project to watch the apiary itself for these kinds of non-standard events (whether by day or night).

Really appreciate your questions Mycroft, you clearly are seeing the big picture! Looking forward to keeping you posted as our team keeps improving the EyesOnHives offering.

Cheers,

Kelton
 

·
Registered
Joined
·
61 Posts
Would you like a collaborator? I have experience with API and protocol design.

It sounds like you have several valuable bits that could be split out and made valuable products in their own right.

For example, the camera is determining "did anything interesting happen" and only uploading "interesting" stuff. Fantastic. This algorithm has broad application for home security; if you could configure it to upload video to an arbitrary server, people could pay to upload video to your cloud service, or to their own home server. I personally am a fan of keeping video in my own home server, or in my own colocated servers. I've seen a lot of Kickstarter projects that look great, and then "oh, but they lock me into using the Cloud..." and I turn away. With this sort of service, you could quickly check who was snooping around your house during the day, and who was the drug addict that rooted through your freezer and stole the pizza.

How often will I look back at 7 years of data? Possibly never. The point of that kind of data storage is, when you need it, you need it. It isn't predictable. But recording only "interesting" things is good enough, doesn't have to be continous feed. I like the way your approach saves bandwidth. But I want 7 years of data on my own server, not chewing up Cloud fees every month.

For the API, something simple like the Slack API is good; simple authentication. Then just some calls to dump the raw data in CSV format. Or even better, S-expressions. And some way to download the video snippets, maybe even by the API. Then people can do their own backup, and the 14 day backlog remains.

I'm not a fanatic open sourcer; people have to eat. But I'm willing to work with you on this. Send over whatever NDA you need to feel comfortable, and I can tell you more about the projects I've been involved with.

For hive theft, with the Android app, possibly you'd even be able to catch the theft in realtime!
 

·
Registered
Joined
·
61 Posts
Another thing; does your analytics detect a swarm leaving or arriving? It would be great to detect when my hives swarm, and also keep an eye on my swarm traps.
 

·
Registered
Joined
·
10 Posts
Discussion Starter · #14 ·
That's very cool, and thank you for your feedback and generous offer Mycroft.

My focus has been on the classification algorithms, my co-founder Jon leads web dev and server infrastructure. I will share our conversation with him and get his take. We have implemented some of the features you mentioned as I'm often pulling our datasets into Matlab before we implement efficient code for the servers to run.

Regarding swarming, so far we have been able to see the different behavior patterns between orientation activity and robbing, but of the 15 hives being monitored so far, we haven't actually had an instance of swarming to classify. Hoping that will change in the new year as we add another 30 hives.

It sounds like we share similar enthusiasm for the potential of this approach, happy you found us and thanks again!
 

·
Registered
Joined
·
61 Posts
Sure. Perhaps we can schedule a Skype conference call between you me and Jon.

For instance, I'm thinking of a portable generator and wifi router to make this work for migratory pollinators with hundreds of hives. Having a server in the truck to store video for upload later would be handy.

The future is endless for this.

I mainly use Lisp, not Matlab. But the stuff I propose doesn't involve the analytics stuff, just the webdev and system level stuff.
 
1 - 15 of 15 Posts
Top