>We need a cost-effective, sensitive detection mechanism for the early detection of Varroa (which we currently lack in Newfoundland & Labrador, as most of you probably know). Timely detection may allow eradication rather than establishment of Varroa. Good luck, eh?<
If you're willing to put in some time calibrating our sensor system, I can help with that problem.
For 15 years we've been working on colony sounds - and no, it's not just the simple frequency shifts and overall noise of the European copy-cats. We were nominated for a National Innovation Award by the Army some years ago for our work with bees and sound.
Our tech barrier was and has been hardware. In 2012, we built a dozen acoustic scanners using Raspberry Pi and demonstrated it at the Vermont meeting of EAS. By the next summer, we had competitors from Europe. Our scanners worked, but they were too big, too expensive, and broke when someone drove over them with a bee truck.
We tried laptops, tablets, phones - all too slow or too cumbersome. The speed of data processing on all but a top-end laptop was too slow. Tablets took 20 minutes or more. Phones stalled out. We're a small group of scientists, not a manufacturing company. The V.C.s/Angels were not going to invest in gearing up manufacturing equipment for such a radical technology, just to get it to market. The barrier was the hardware.
Two years ago, that wall fell when top-end Smart Phones added Facial Recognition to their security options. Facial Recognition uses A.I.
Our acoustic system uses A.I. - it's the sound counterpart of visual A.I. Both pattern recognition systems require fast processors and robust storage capability. With the iPhone and Android 8 phones, that capability became available.
Please note, our A.I. systems can 'learn' patterns. We've had very good results for varroa in Alberta and SE U.S. Bees from different geographical locations tend to have the bee version of an accent. For best performance, the program may need to be trained to local bees. We've made that easy, if local beekeepers, students, scientists are willing to do the field work.
Two years ago, we started on a Smart Phone acoustic app for version 8 and newer Smart Phones. Our acoustic system is patented in U.S. and in Canada. The App was 'supposed' to be delivered Jan of 2017. It's now 27 months, and we finally have a reasonably complete app.
We made a smart move and hired an expert in app design. As such, the resultant app is easy to use. The app works regardless of whether cell service is available. For data handling, everything is transferred to the Cloud at the click of one button - when the user gets near a fast wireless hot spot/router. The upload system can handle over 100,000 simultaneous users. Recordings, analysis results, and all other data is stored in a secure, expandable, cloud-based database.
As it stands at the moment, the user inserts the microphone end of the phone into the entrance of a hive or an external microphone. Push one button, and everything starts. The sound recording takes 30 seconds. Data processing by A.I. takes about 12-15 seconds. While that's going on, the phone captures the date, time, and GPS location, type of phone, operating system, and brand of phone. The app automatically runs multiple analyses of the sound recording and displays the results as soon as the analyses are done. It takes less than a minute for everything - recording, analysis, analysis results, with the option of then visually inspecting the hive. Hive inspections (e.g., alcohol wash, stick traps, visual inspections) will be used to confirm or reject analysis results and categorize the recordings for A.I. re-training.
Everything is stored automatically on the phone. Everything, including recordings and inspection reports, uploads to a cloud site with the push of one button. We get everything from every user, anywhere in the world.The basic app from beginning to submission of all of the data requires a few, single button clicks.
Because we have no control over the hardware used by phone manufacturers, nor the genetic makeup of the honey bees in any area, we expect to need to re-train our analysis systems based on actual recordings and inspection reports submitted by users.
The system does not 'learn' on the phone - it's not self-learning like Alexa or Siri or Google Assistants. It could be, but that takes time and a LOT more money. Amazon or Google could quickly fix that problem.
In the meantime, as we receive uploaded recordings, analysis results, and inspection reports, we will dump all of the recordings with correct analysis reports into one set of data, all of the incorrect ones into another, and re-train the A.I.s. As we progress, the A.I.s are proven to get better.
By 2012, our acoustic system had an 86-98% accuracy for multiple pest, disease, and chemical exposure factors - when the recordings were taken using high-end digital recorders and bench-top PCs and Work Stations.
We don't expect the app to work properly on all phones in all areas for all factors on first use. We expect the app to dial in as we obtain sufficient numbers of recordings/analysis/inspections to re-train the app for specific user groups, areas, hardware, as needed.
During our beta testing, we found some factors that needed to be re-trained, varroa being one. We found some others that worked surprisingly well on first try (e.g., American Foul Brood). Foul Brood worked surprisingly well on both iPhone and Android, and it may be more sensitive than even Elisa tests (but we shall see).
We have been in a private, beta-testing mode for 2 years. That has allowed us to find and fix a variety of issues. We had to see if it worked on Smart Phones - and it does; but it clearly needs re-calibration for some factors. That can only be accomplished by experienced and dedicated users. Using the app and not sending back the recordings, analyses results and inspections does no one any good.
We're planning a 2019 public launch by summer of the App of iPhone, Android, and Windows platforms. We're looking for groups who have the interest and ability to contribute to accurate re-calibrations
We've done some work in NZ and Australia - each of those countries have bees that sound different from N. American bees. Much like English, that sounds different than U.S., Canada, U.K., New Zealand, and Australia, bees have dialects or accents.
When the App is properly dialed in, we know that we can even discriminate race of bee in the US, as well as find commonalities for the major pest and disease issues that over-ride the subspecies variations in bee sounds.
The App is ready for next stage of testing. We will need testers - especially those who can provide good data for refining the training to various pests and diseases. We're looking for researchers with varroa farms, people studying various pests, commercial folks who have small hive beetle problems, queen breeders, etc.
We have done the R&D. We know what the app can do. We've spent considerable time making the App ready for distribution, so that we can achieve it's full potential. We've got the entire Cloud data system in place.
What we do not have is funding for our geek squad to grab all of the data streaming to us and use it to re-train the A.I.s. We're going to need to pay our resident geek, who built all of the custom A.I.s. I'm looking for funds for his salary for about a year to see to all of the re-training and updating.
In the tech transfer world, we's say that we're posed to enter the 'valley of death'. After 45 years of doing strange and weird things with bees, like landscape pollution monitoring, conditioning bees to find landmine, developing and building decimeter resolution LIDAR instruments to map forager bee locations in 3-D space, this may be the capstone.
Immediate, on site, detection and grading of bee pest and disease problems with a Smart Phone - no need to open the hive expect to verify the analysis results. As it becomes more accurate, we might even be able to drop the last (inspection) step. In theory, we may be able to say - all of those colonies are fine, no need to inspect. All of these colonies are really sick or dead - time to clean-up and get rid of them. And these remaining colonies should be inspected to see what's going wrong.
Since all of our data goes to the cloud, we have versions of the app that simply report to the user at the hive (Hobby level), a version that provides daily summary reports of all hives scanned (appears on the phone), and a version for commercial beekeepers that will provide to the owner or manager daily reports from all crew members, all hives scanned, for all locations in a the bee operation.
We've put our blood, sweat, and tears into this for over a decade. Now we need to fine tune it.
Jerry
P.S. The App is not available to the public, yet. Early release access is on a case-by-case basis. I don't need nor want individual - I've got 1-2 hives at this time.
I do want to talk to anyone who can provide reliable inspection data in concert with recordings and analyses from multiple colonies (e.g, varroa monitoring, small hive beetle numbers, queen right versus queenless, etc.). Participants will benefit from early participation by being able to help shape the look and feel of the app, get their colony data into the calibration and training early on. We will require an NDA. We've already got competitors. The good news - they've also found value in acoustics. The bad news, many seem more interested in selling their apps than whether it actually works or benefits beekeepers and the beekeeping industry.
Personally, I believe that this has the potential to revolutionize beekeeping and pest and disease detection and tracking. Everyone with a decent Smart Phone could use this. We'd get daily reports from all over. Varroa? What if we knew the day and week that varroa levels started climbing is any reporting area? I can imagine the equivalent of the CDC human influenza outbreak mapping and tracking for the US and Canada. Something new appears, we'd now where and when, maybe even find ground zero. Commercial beekeepers, add one crew member - someone responsible with a vehicle and driver's license. Have that runner go ahead of the crews. Each evening a report of healthy, sick, and suspect hives by hive and location would be delivered to the manager/owner. Crews would be sent to those hives that need attention at locations where the colonies need attention.
This is the primary reason I've failed at retirement. I turn 73 in a few days. I am painfully aware of the count-down clock running. Before my mind or health goes, I want to see this happen. We've known since early 2000s what can be done.
Technology limitations of speed, storage, size, cost were barriers; just as battery technology governed when electric cars would hit the market. That's been promised since I was first old enough to read Popular Mechanics. The electric car for the mass market was always just around the corner. Electric cars, even performance electric cars, are now here. Self-driving cars are close. Combines track position, crop yields, and self-drive. I've gone from my dad's phasing out of horses and going to tractors when I was young to a truck and car that talk to me, warn of hazards.
Dick Tracy's watch was every kids dream. Mine dream is to provide the equivalent of Dick's watch to bees and beekeeping to every beekeeper with a Smart Phone-A Tricorder for bees. I'm convinced this can happen. It will be a disruptive technology. Whether it goes the way of the Dodo or becomes the Bee Management Tesla remains to be seen. It's going to depend on people like those on this list.
***********************************************
The BEE-L mailing list is powered by L-Soft's renowned
LISTSERV(R) list management software. For more information, go to:
http://www.lsoft.com/LISTSERV-powered.html
|
