Connected beehive as MSc thesis project

Hello everyone, I am currently doing my masters in robotics, and my thesis is about making a beehive IOT monitoring system, wanted to ask for any suggestions for the project and what to add to it.


Dear @hamze123,

welcome to the Hiveeyes community, and good luck for your masters thesis.

Looking at the backend side, Rui Pinto also used fragments of the Hiveeyes infrastructure to whip up his MSc thesis, see Mosquitto authentication fails when using password made of digits · Issue #15 · daq-tools/kotori · GitHub.

However, that’s just a small part of the whole system, I just wanted to share some insights about backend and telemetry infrastructure, also offering any support you may need on this matter.

For an index of the most valuable material and spots on the community forum, please enjoy Awesome Hiveeyes -- A list of curated resources, but please also don’t consider this exhaustive. When in doubt, please ask.

With kind regards,

1 Like

hello @Andreas

Thank you so much for having me and thank you for the reply. I will be sure to look at the repositories and sources you sent me and will ask any questions if needed. The project is tailored more towards vision and the use of sensors. The backend, in terms of login and authentication, will be added as a bonus when time permits.

Thank you again.
Regards, Hamze.

May you share some thoughts and plans with us? We have used – in the most cases – a scale, temperature and humidity sensors, sometimes IR cams and did some steps toward sound analysis.

Sound analysis is actually something I was interested in doing, specifically for detecting the presence of a queen bee, There are also some sources on counting bees in images with a high density of bees and how it could be done, Here are some interesting papers you can look at for both topics
sound analysis: Analyzing the Beehive’s Sound to Monitor the Presence of the Queen Bee | IEEE Conference Publication | IEEE Xplore
counting bees: A method for automatic honey bees detection and counting from images with high density of bees | IEEE Conference Publication | IEEE Xplore

Ive also messaged bee keepers about some featuers they might find useful, and one was mite detection, Here is a paper on that as well: AgriEngineering | Free Full-Text | Detection of Varroa destructor Infestation of Honeybees Based on Segmentation and Object Detection Convolutional Neural Networks There are alot of datasets available on kaggle: that could prove useful, There are so many ideas that can be implemented into the system using vision a Raspberry pi with cameras, possibly

One thing to keep in mind for the Raspberry Pi is that it doesn’t have the best processing, there are more powerful systems, like Nivida Jetsons, which have Cuda and much better processing, however, you can also use TPUs for that (tensor processing units): USB Accelerator | Coral

in terms on sensors, im not sure tbh, as the sensors you sent were mostly what I was thinking of using, I was thinking a fire sensor as well outside the bee hive to detect wild fires, air quality sensors maybe other than that, im not really sure still researching

We did some research in Sound Visualization that may be helpfull. We did also some FFTs with sound data from hive with queen vs queenless colonies, and it is surprising easy to see the queenless: Sound Samples and Basic Analysis Hive with Queen vs. Queenless . So I think this should be doable!

From a practical beekeeper perspective – btw are you a beekeepter also? – this is not helpfull I fear, at least for Europen colonies. All honey bee colonies have varroa mites in the hive. If you can see them on the bees it is in the most cases too late. The mites hide in the brood and between thorax and abdomen so they are in the most cases invisible. Varroa mites can only be seen if the damage threshold is very high!

Bee colonies are often far away from a power plug and also internet connection. So for testing a Raspberry Pi is ok, but perhaps it is more practical to send raw data to a server and make there the magic. But I fear for elaborated AI applications this will not work.

It may depend on the area of beekeeping, fires are no danger in Europe (till now, may vary with climate change). Perhaps a view at the BME688, Gas-Sensor mit Artificial Intelligence (AI) may be worth, in and outside the hive.

Unfortunately I’m not a beekeeper, this project was proposed by my professor and I’m just currently doing some research on what would be possible as my professor wants “something new” implemented, as for varroa mites there might be a way to actually detect them early I’m just saying this in theory but even in medical applications, machine learning models we’re able to detect cancer cells earlier than humans, when they are basically barely noticeable to human eye, of course the missing key here to finding out if it’s possible is data. Thank you for mentioning that as I can actually write it down in the project proposal, raspberry pi is just meant for quick prototyping, and unfortunately all my testing will be on simulation anything physical will prob be done as prototype but won’t actually involve honey bees for safety issues. For sound analysis is there any available data of recorded sound samples, Visualisation is more for proof of concept however will need large data if it to be implemented into an ML model and I think my professor would like the idea even if it’s just a concept.

Also before I write down what you said about mites, is it fine if I reference your comment? Else I can just find a different source, let me know

Varroa mites are always in every hive, there is no hive without varroa since decades. So the cancer example is not appropriate. Varroa mites on top of bees is like seeing the death on the face of a sick person. It’s really the pre-final step before the final hive colapse.

You can find some snippets at Sound Samples and Basic Analysis Hive with Queen vs. Queenless - #17 by clemens

and there are no treatments for varroa mites ?

There are but you can only reduce, never extinct the mites!


what other things might also pose threat to honeybees ? i heard yellow legged hornets, maybe.

Before diving into building a beekeeping tool, consider researching beekeeping itself. My professors always emphasized that understanding the bees’ food sources, the plants they rely on, is crucial for truly knowing the insects. The deeper you delve into their ecosystem, the more this becomes apparent.

Of course, another approach would be to develop a questionnaire and directly ask beekeepers about their needs.

Think outside the box! Perhaps instead of focusing solely on threat detection, consider tools that highlight positive impacts.

You mentioned yellow-legged hornets?

If you’re interested in detecting Vespa velutina specifically, there’s a wealth of existing research and projects. Check out VespAI and VespAlert, which often utilize Yolo for real-time detection.

PS: Formulating a hypothesis and research question is one of the hardest parts! :) Spending here more time will help you definitely later on.


in the process of doing that Yes, after I do this semester’s exam, I will delve more into beekeeping itself over the summer, hopefully. I am just in process of writing the proposal at the moment, as that is due on April 23. thank you so much for your input and I will look into sources provided :)
I have read a bit more into bee behavior as well and one of the things i found out is that social bees actually make up less than 10% of the bee’s population and up to 80% are actually solitary bees. I also read about social parasites as well, which was very interesting, but i will probably write a short section of the practices of bee keeping in the background research, but dont have much time to write in details at the moment.


@hamze123, let’s hear about your progress and/or (prelim?) results! :)


It is exam season so not much, I have submitted the project proposal and then hopefully after exams are over I’m able to start working on it during summer.

1 Like