DODonHopkins·Aug 19, 2023
Video from the livestreamed web camera at the legendary BrainWash cafe/laundromat in San Francisco was collected and put into a public dataset and used for training, then later "This data was removed from access at the request of the depositor."
As facial recognition draws scrutiny nationwide, Stanford research raises questions closer to home:
https://stanforddaily.com/2020/07/28/as-facial-recognition-d...
>The camera didn’t attract much attention.
>It was perched in a corner of San Francisco’s BrainWash cafe and offered an unobstructed sweep of the room. That’s probably why the owners placed it there, and also likely what made it so appealing to the authors of a 2015 Stanford paper on using artificial intelligence (AI) to pick out people in crowded scenes.
>The camera streamed its footage live on the internet, where it was collected by t
Read the rest
he Stanford researchers to train and validate their algorithm’s effectiveness. Clips from the database, preserved on YouTube, show BrainWash customers milling about, ordering coffee or muffins.
>None of them look like they know they’re going to become the raw material for a computer vision dataset that will be downloaded dozens of times around the globe, appearing in Chinese research linked to human rights abuses and sparking national concern over the power and potential of computer vision and facial recognition software.
>A month after it was mentioned in the Financial Times on privacy concerns over computer vision, Stanford removed the Brainwash dataset from its public online archives “at the request of the depositor.”
https://searchworks.stanford.edu/view/sx925dc9385
>Brainwash dataset
>Creators/Contributors: Author: Stewart, Russell
>Abstract:
Brainwash dataset for face detection. See implementation of algorithms at github.com/russell91/tensorbox.
>Subject: Stanford Department of Computer Science
>Genre: Dataset
>Preferred Citation:
Stewart, Russell. (2015). Brainwash dataset. Stanford Digital Repository. Available at: http://purl.stanford.edu/sx925dc9385
>Location
https://purl.stanford.edu/sx925dc9385
>Access conditions
>Use and reproduction: This data was removed from access at the request of the depositor.
End to end people detection in crowded scenes - results on test set
https://www.youtube.com/watch?v=QeWl0h3kQ24
>Supplementary material for http://arxiv.org/abs/1506.04878
The dataset even included some (silent) footage of Tony Sparks Open Mic Comedy Night! I would usually sit by the door to the laundromat where the plug was, to the right and out of sight of the camera, far away from the stage the camera was pointing at, to avoid being taunted by the crazy comedians.
"The Godfather Of Comedy" Discusses The City's Longest-Running Open-Mic:
https://hoodline.com/2015/08/the-godfather-of-comedy-discuss...
>“We’ve had some things that are really funny or weird, like a guy who climbed into a box and did his act from there,” Sparks remarked. “[Another] guy wore a green suit, all green, face covered up and everything. We had a guy who did it naked. We had a girl who pulled her top off. Sometimes some of these people just need someone to look at them and talk to them, you know?”