[maxmunzel]

Hey, author here — thanks for posting it again!

As of 2025 the pigeons are back and version 2 with aiming and a microphone array for target acquisition is still work-in-progress. I’m heading to bed now, but I’ll happily answer any questions tomorrow.

[guest__user]

I am using a raspberry pi and a two axis turret with a pump and reservoir from a car windshield sprayer. right now it's just using differences in pixels to detect movement, by it it's keeping the balcony pigeon free! (a few false positives when a bumble bee flies too close to the camera though) I also have a yolo model version but haven't implemented it yet !

[maxmunzel]

Cool! Do you have the STLs of the turret online somewhere?

My setup also had its occasional false positives but that’s not a problem if all you do is spray a bit of water around. I also looked into yolo, but I couldn’t find any good datasets of pigeons. Nowadays that’s no problem anymore as you can just use models like meta’s segment anything that do zero-shot bounding boxes… if you have the GPU to run it.

[Taek]

The introduction of GPT 4o probably gives you a much simpler path to identifying the location of pigeons right? Is that something you have looked into?

[maxmunzel]

Not yet, but I used VLMs to read out water meters with mixed results. It is definitely the easiest approach to prototype. For my balcony, all vision based approaches are limited by viewing angles that don’t expose any neighbors.

But even with an llm backbone you’d still need some setup to detect when to query the llm. And then you already have a low-false-negative pigeon detector that may be sufficient for your use case.