Live data from Hacker News

Raspberry Pi for Kill Mosquitoes by Laser

preprints.org

11–20 of 262 posts

Re: Raspberry Pi for Kill Mosquitoes by Laser

#12
This work (I think) worked with the mosquitos 30cm away with a servo scanning Pi Camera (1080p) and a 1W laser.

To work in the real world - cover a whole room or terrace - presumably a much higher resolution camera (or much faster scanning system) would be required. Even a 1W laser is dangerous to eyesight, if it was being fired at targets mingling with people.

The system could be mounted on small drones that would patrol larger areas - but the idea of robotic drones armed with lasers roaming around is beginning to sound worse than the mosquitos.

Re: Raspberry Pi for Kill Mosquitoes by Laser

#13

Looks like a fun project but the results did not look promising. Out of 100 mosquitos the best algorithm would’ve taken out less than 10.

In a laboratory environment with contrasting background—something they admit isn’t ideal, I assume for real world conditions.

Let’s say the laboratory conditions are real world. To your point, killing ~10% of mosquitos is hardly ideal, but using lasers and computers to do it is cool.

6/10. Want to try.

Re: Raspberry Pi for Kill Mosquitoes by Laser

#15
An interesting future application for computer vision, but until both hardware and software increase in speed and precision by several orders of magnitude, this is not feasible. Current neural object detection just isn't there yet. Not to mention the dangers of shooting a powerful 1W laser out in the open...

Re: Raspberry Pi for Kill Mosquitoes by Laser

#16

This work (I think) worked with the mosquitos 30cm away with a servo scanning Pi Camera (1080p) and a 1W laser. To work in the real world - cover a whole room or terrace - presumably a much higher resolution camera (or much faster scanning system) would be required. Even a 1W laser is dangerous to eyesight, if it was being fired at targets mingling with people. The system could be mounted on small drones that would p…

I agree, there needs to be multiple orders of magnitude improvements in HW&SW, as well as domain specific dataset development for this to work at all. I can imagine using a very accurate image segmentation algo to analyze the background and thereby prevent shooting lasers at vulnerable targets, but it's still hard to see how this is a good idea.

A 1W laser is even more dangerous than you say. It is powerful enough to start wood fires.

Post reply on HN