Naive me (without bee-knowledge) was already excited for a modified barcode engine which can identify individual bees just from scanning their stripes... :(
Me three. I was hoping for a modifidable stripe configuration to carry data.
A swarm of bees being a dancing-byte-string-of-sorts travelling across a landscape.
I've not yet read to the end so apologies if this idea is covered already, but a ctrl+F for "colour" (and "color") yields nothing so I'll comment away:
> "If you’re thinking ahead you’ll realise that the maximum number of bees they could therefore simultaneously study was 2048. That’s about 1/25th of a very strong colony at the peak of the season, or the number of bees covering both sides of a two-thirds full frame of sealed brood."
For a system you're total controlling (in that they created their own QR-inspired format based on the size that can fit on a bee), would it not be simple to use colours to expand the range?
For example just having red+white codes as well as black+white ones you double from 2048 to 4096 bees you can label. Or is there some issue that bees wouldn't like other colours, or some other problem I'm not foreseeing?
And I don't know how far it could extend, but I guess even 25 shades of various colours could be different enough from each other for the camera scanning them to differentiate?
The room was only illuminated with infrared light which means that you can’t easily discriminate color + taking and processing photos in color would probably take more time and storage
Higher speed and scientific cameras tend to be black and white to increase sensitivity, speed and pixel count. Looks like the camera they used is also B&W in this case.
This is why colors wouldn't be practical. Also, when you're tracking things in high-ish speeds, motion blur can shift colors because of Bayer filtering, when you're looking from that distance. Yes, demosaicing will remove the shift, but the resulting color might be different.
So, B&W is the safest bet with minimum technical complexity.
2D markers that use color to encode extra bits of information do exist actually, see for example http://josephdegol.com/docs/ChromaTag_ICCV17_Paper.pdf. So it's not a bad idea. But as other commenters mentioned, for this application it might not be a good fit if the researchers are limited to using black and white cameras (understandably - there's a hardware tradeoff to capturing color information in terms of speed and resolution). More generally, color information is harder to capture robustly under varying illumination conditions, not to mention the accuracy of color printing, so these color markers haven't really caught on to my knowledge.
A bit of a tangent, but reminds me of this article, RenderGAN [1]. The main point of the article was about using 3D rendering + GAN to generate convincing synthetic data, and their application was creating training data to detect "barcodes" [2] attached to bees.
It's a similar premise and method than Apple's SUGAN paper [3], that came out around the same time, to much more fanfare (it won CVPR best paper award iirc). I think the impact of renderGAN was a bit limited by the narrow scope of their application (even narrower than Apple's eye tracker application) and the fact that the generator had some handcrafted layers that didn't really generalize well outside the black and white marker application.
[2] Technically not a barcode since it doesn't have bars, it's a circular 2D marker, but close enough. For that matter the QR-like codes aren't really "bar"codes either, but that's just pedantry.
I've pasted QR codes on the backs of bees before! Unfortunately we started this project a week before COVID so we didn't get to see it through, but the actual gluing is easier than you might think. Doesn't look like they used it here, but we planned on using James Crall's BEEtag repo [0].
I'm working on something to let you measure animal activity without pasting QRs [1]. I've been running a casual study of my bird feeders since September, and my system will be field-deployed by a few labs this summer. My background is in pollination ecology, so bee/pollinator tracking is a top priority. If you want to study your own backyard, you can use polliOS with your own IP cameras. Targeting March 1 for beta release, but you can submit your email to get notified.
I get it, as which is harmless too. However to me tagging a bee for the outfit itself kind of destroys the natural element of an bee. Is it really necessary?
You could argue that a beehive is a dedicated home that you define for the queen, so the wild in wildlife is already lost.
I'm aware they do the same for birds with tags on their talons which I still feel is unfair as we don't tag humans with barcodes (yet).. maybe by other means but, I dunno.
16 comments
[ 4.9 ms ] story [ 45.0 ms ] threadI will stop now.
> "If you’re thinking ahead you’ll realise that the maximum number of bees they could therefore simultaneously study was 2048. That’s about 1/25th of a very strong colony at the peak of the season, or the number of bees covering both sides of a two-thirds full frame of sealed brood."
For a system you're total controlling (in that they created their own QR-inspired format based on the size that can fit on a bee), would it not be simple to use colours to expand the range?
For example just having red+white codes as well as black+white ones you double from 2048 to 4096 bees you can label. Or is there some issue that bees wouldn't like other colours, or some other problem I'm not foreseeing?
And I don't know how far it could extend, but I guess even 25 shades of various colours could be different enough from each other for the camera scanning them to differentiate?
This is why colors wouldn't be practical. Also, when you're tracking things in high-ish speeds, motion blur can shift colors because of Bayer filtering, when you're looking from that distance. Yes, demosaicing will remove the shift, but the resulting color might be different.
So, B&W is the safest bet with minimum technical complexity.
Addenda: Oh, I missed the IR lighting, too. Color IR is hard. These guys do it: https://kolarivision.com/articles
It's a similar premise and method than Apple's SUGAN paper [3], that came out around the same time, to much more fanfare (it won CVPR best paper award iirc). I think the impact of renderGAN was a bit limited by the narrow scope of their application (even narrower than Apple's eye tracker application) and the fact that the generator had some handcrafted layers that didn't really generalize well outside the black and white marker application.
[1] https://www.frontiersin.org/articles/10.3389/frobt.2018.0006...
[2] Technically not a barcode since it doesn't have bars, it's a circular 2D marker, but close enough. For that matter the QR-like codes aren't really "bar"codes either, but that's just pedantry.
[3] https://arxiv.org/abs/1612.07828
I'm working on something to let you measure animal activity without pasting QRs [1]. I've been running a casual study of my bird feeders since September, and my system will be field-deployed by a few labs this summer. My background is in pollination ecology, so bee/pollinator tracking is a top priority. If you want to study your own backyard, you can use polliOS with your own IP cameras. Targeting March 1 for beta release, but you can submit your email to get notified.
[0] https://github.com/jamescrall/BEEtag [1] https://polli.ai
I get it, as which is harmless too. However to me tagging a bee for the outfit itself kind of destroys the natural element of an bee. Is it really necessary?
You could argue that a beehive is a dedicated home that you define for the queen, so the wild in wildlife is already lost.
I'm aware they do the same for birds with tags on their talons which I still feel is unfair as we don't tag humans with barcodes (yet).. maybe by other means but, I dunno.
Where we are going with this world?