Does it work? Is this like drawing a mustache on a kid? I want to believe it couldn't work, but if they're using AI to detect trucks then who the hell knows?

[–] lokalhorst@feddit.org 10 points 2 days ago (1 children)

It depends largely on which sensors the AI was trained on. Could be a combination of RGB+I, LiDAR etc. If only RGB, this is problematic for object detection.

[–] Hubi@feddit.org 7 points 2 days ago

Most of these longer range drones utilize infrared vision. The camo is not gonna do shit. There are generally a lot of misconceptions around these AI drones in the Russian army, the main one being the alleged face tracking capabilities from that one video.

[–] fizzle@quokk.au 9 points 2 days ago (2 children)

I have no idea, but I have some supposition: no it won't work.

Like others have said warships used to do something similar.

The naive idea is a WWII concept: that the hard lines make it more difficult to see your heading from a distance.

The problem is drones don't "see" the way people do.

[–] AngryCommieKender@lemmy.world 6 points 2 days ago

Also beyond not seeing the way people do, the drones are constantly doing range finding and tracking, so they can adjust their trajectory on the fly. The WWI shells and torpedos didn't do that.

[–] teyrnon@sh.itjust.works 3 points 2 days ago

We could trick the drones, they would be reprogrammed to be untripped and maybe a couple months or less I'm sure.

[–] Clent@lemmy.dbzer0.com 9 points 2 days ago* (last edited 2 days ago)

Looks like two different attempts could be testing designs.

[–] clav64@lemmy.world 3 points 2 days ago (1 children)

It makes it harder to see the edges and so can't determine the objects shape, and when moving potentially how fast or in which precise direction it is moving in.

LIDAR on the other hand doesn't care for dazzle paint.

[–] ohulancutash@feddit.uk 1 points 1 day ago (1 children)

Until someone gets their paws on vantablack

[–] Tar_alcaran@sh.itjust.works 2 points 1 day ago (1 children)

"if return=0, set target_true"

[–] clav64@lemmy.world 1 points 1 day ago

Save the AI tokens and just use straight logic.

[–] gegil@sopuli.xyz 4 points 2 days ago (1 children)

I remember when facial recognition technology began to be mass deployed, it was possible to trick those algorithms with special makeup. The idea of this camouflage is to trick ai drones in the same way. But it will not work with modern ai technologies, or if drone uses other targeting system like thermal cameras, or if drone simply being operated by person.

[–] teyrnon@sh.itjust.works 3 points 2 days ago

Around like the zeros to 2010 at least the facial recognition was pretty lacking. A band aid on the face could defeat it.

Nowadays though it's gotten better and they do other stuff by like analyzing your gait, obviously the eyes if you get close enough, speech patterns.

Basically we just have to stop these people that control the information because they have too much of it. But that is another story.

But not long after 2010 I seen recall a bunch of fugitives getting caught and I am willing to bet it was because of facial recognition.