Scraping Tinder selfies to make a dataset for AI experiments

Re:Scraping Tinder selfies to make a dataset for AI experiments by danso on Saturday April 29, 2017 @04:31PM (#14228590)

International laws are all different, but American laws are generally among the most permissive. For example, newspapers don't have to get consent forms for everyone who happens to appear in a photo of a public venue. I don't think Google was legally obligated to blur faces in its Street View coverage of America, but it was likely the extremely wise thing to do (nothing stopping government from passing new laws) [0]. But documentaries do -- I'm honestly not sure if that is legally the case or if it's done out of precaution, obviously the majority of TV news and newspaper photos don't blur random people's faces. And for things like the "Ali G Show" [1], you can bet all of those people who were unblurred had signed a release before entering that bar.

In contrast, I know that in the EU, architects have copyright on their work, which means you can't photograph the Eiffel Tower at night and share the photo, legally, without consent [2]

As to how you know if your photo is being used commercially...obviously there isn't a push-notification service (yet) that sends an alert to your inbox. You would just have to notice the commercial usage, as what happened when the teenager who was captured in a CC-licensed photo on Flickr found herself in a Virgin Mobile campaign

[0] https://www.cnet.com/news/google-begins-blurring-faces-in-st... (warning, autoplay video)

[1] https://www.youtube.com/watch?v=Vb3IMTJjzfo

http://www.thefrisky.com/2009-07-15/how-sacha-baron-cohen-ke...

[2] http://www.politico.eu/article/banned-taking-pictures-of-the...

[ Parent | on HN ]

Re:Scraping Tinder selfies to make a dataset for AI experiments by ghaff on Saturday April 29, 2017 @05:28PM (#14228857)

>You would just have to notice the commercial usage, as what happened when the teenager who was captured in a CC-licensed photo on Flickr found herself in a Virgin Mobile campaign

The issue in that case was (IANAL) nothing to do with the CC licensing but with the fact that the photo was used for a commercial (i.e. marketing/advertising) purpose. Those require a model release form, hence the requirement by microstock sites for model releases.

Virgin Mobile messed up in that they used a photo for a marketing campaign that didn't have a model release form either directly or through an agency that had vetted it.

Editorial use in the US at least doesn't require this even though it's commercial use in the sense that a for-profit entity is using the photo to illustrate a story.

[ Parent | on HN ]

Re:Scraping Tinder selfies to make a dataset for AI experiments by danso on Saturday April 29, 2017 @05:45PM (#14228941)

I didn't mean to imply that the CC licensing was at fault, though IIRC the girl and her family did originally include the CC Corporation as a defendant [0] before dropping them from the suit. I was merely stating the fact that just because the photographer has made their photo freely available under CC does not mean the subject has given consent for their image to be commercially used. Virgin Mobile screwed up into thinking that's what CC meant.

[0] http://fairuse.stanford.edu/case/chang-v-virgin-mobile-creat...

[ Parent | on HN ]

Re:Scraping Tinder selfies to make a dataset for AI experiments by ghaff on Saturday April 29, 2017 @05:52PM (#14228978)

Absolutely right. I actually wrote about this at the time. [1] That CC has commercial/non-commercial variants probably add further to the confusion even if they didn't play into this case. As does the fact that the colloquial meaning of commercial is different from the meaning of commercial that's relevant to publicity rights.

[1] https://www.cnet.com/uk/news/more-commercial-creative-common... (Video autoplay, grr.)

[ Parent | on HN ]

Re:Scraping Tinder selfies to make a dataset for AI experiments by devrandomguy on Saturday April 29, 2017 @04:41PM (#14228632)

I guess, I just have very little faith in my ability to get personal justice versus a foreign corporation. The value of the dataset, when applied to an unethical but lucrative purpose, could easily outgrow the cost of getting caught. Especially if the operation is well isolated from the parent company.

It is not the possibility that I might appear in an ad that concerns me, but rather, that the set of photos that contain me in them, could be used correlate way more information than I would ever consent to releasing voluntarily. I would not even notice that happening.

[ Parent | on HN ]