zlacker

[parent] [thread] 10 comments
1. Pxtl+(OP)[view] [source] 2020-06-24 15:00:13
Intresting and related, a team made a neat "face depixelizer" that takes a pixelated image and uses machine learning to generate a face that should match the pixelated image.

What's hilarious is that it makes faces that look nothing like the original high-resolution images.

https://twitter.com/Chicken3gg/status/1274314622447820801

replies(5): >>mywitt+q >>jacque+31 >>danso+t3 >>nfrmat+I4 >>2038AD+Yp2
2. mywitt+q[view] [source] 2020-06-24 15:02:17
>>Pxtl+(OP)
I wonder if this is trained on the same, or similar, datasets.
replies(1): >>jcims+qx1
3. jacque+31[view] [source] 2020-06-24 15:05:08
>>Pxtl+(OP)
That should be called a face generator, not a depixelizer.
replies(1): >>Polyla+1G1
4. danso+t3[view] [source] 2020-06-24 15:14:38
>>Pxtl+(OP)
What's sad is that a tech entrepreneur will definitely add that feature and sell it to law enforcement agencies that believe in CSI magic: https://www.youtube.com/watch?v=Vxq9yj2pVWk
replies(1): >>barrke+be
5. nfrmat+I4[view] [source] 2020-06-24 15:19:40
>>Pxtl+(OP)
Interesting... Neat... Hilarious... In light of the submission and the comment you're responding to, these are not the words I would choose.

I think there's genuine cause for concern here, especially if technologies like these are candidates for inclusion in any real law enforcement decision-making.

◧◩
6. barrke+be[view] [source] [discussion] 2020-06-24 15:56:35
>>danso+t3
And another entrepreneur can add a feature to generate 10 different faces which match the same pixelation, and sell it to the defence.
replies(2): >>emilio+3H >>heavys+7H
◧◩◪
7. emilio+3H[view] [source] [discussion] 2020-06-24 17:42:42
>>barrke+be
A better strategy might be to pixelate a photo of each member of the jury, than de-pixelate it through the same service, and distribute the before and after. Maybe include the judge and prosecutor.
◧◩◪
8. heavys+7H[view] [source] [discussion] 2020-06-24 17:42:49
>>barrke+be
Doubt that many people can afford to hire an expert witness, or hire someone to develop bespoke software for their trial.
◧◩
9. jcims+qx1[view] [source] [discussion] 2020-06-24 22:18:01
>>mywitt+q
One of the underlying models, PULSE, was trained on CelebAHQ, which is likely what the results are mostly white-looking. StyleGAN, which was trained on the much more diverse (but sparse) FFHQ dataset does come up with a much more diverse set of faces[1]...but PULSE couldn't get them to converge very closely on the pixelated subjects...so they went with CelebA [2].

[1] https://github.com/NVlabs/stylegan [2] https://arxiv.org/pdf/2003.03808.pdf (ctrl+f ffhq)

◧◩
10. Polyla+1G1[view] [source] [discussion] 2020-06-24 23:27:00
>>jacque+31
Basically. The faces look plausible but less useful than the original blurred image.
11. 2038AD+Yp2[view] [source] 2020-06-25 07:37:39
>>Pxtl+(OP)
Ironically, if the police had used and followed the face depixelizer then we may not have had the false arrest of a black man - not because of accuracy but because it doesn't produce many black faces
[go to top]