Facemasks work great. There’s a reason the gestapo are using them.
People like the makeup because it looks like a cool cyberpunk future, but the moment a pipeline is designed to compensate for made up faces it will fail. Simple image processing can replace makeup hues and yield a roughly true color unobscured face.
That’s not what the problem is. It only fools systems that just feed the image straight into a race recognition model without trying to correct for it. You don’t need to even retrain the model, just use basic image processing to color swap the tattoo before the model even gets it. The tattoos are easy to isolate computationally.
Facemasks don’t work very well. Makeup is a far better tool.
https://www.vice.com/en/article/researchers-defeated-advanced-facial-recognition-tech-using-makeup/
Study here: https://arxiv.org/pdf/2109.06467
And an example –
Here’s an updated article from 2025 that works by using makeup in a subtle fashion.
https://www.theregister.com/2025/01/15/make_up_thwart_facial_recognition/
Facemasks work great. There’s a reason the gestapo are using them.
People like the makeup because it looks like a cool cyberpunk future, but the moment a pipeline is designed to compensate for made up faces it will fail. Simple image processing can replace makeup hues and yield a roughly true color unobscured face.
Actually its been proven that some face paint in these patterns fools AI from recognizing a face
That’s not what the problem is. It only fools systems that just feed the image straight into a race recognition model without trying to correct for it. You don’t need to even retrain the model, just use basic image processing to color swap the tattoo before the model even gets it. The tattoos are easy to isolate computationally.