So Yann should we as a community just build something as a proof of concept? If we are talking less than 250 USD per month, I am sure we can scrounge up the money for a trial 6 month trial.
James
On Tue, Jun 18, 2019 at 2:59 AM Yann Forget yannfo@gmail.com wrote:
Hi,
Yes, James' pricing doesn't match the actual cost. We do not need to check all images uploaded to Commons, only the suspicious ones (small images without EXIF data). If we check 2,000 images a day (more than enough IMO), that would cost $7 a day, so $210 a month.
Regards, Yann
Le mar. 18 juin 2019 à 01:11, James Salsman jsalsman@gmail.com a écrit :
Google has been offering reverse image search as part of their vision
API:
https://cloud.google.com/vision/docs/internet-detection
The pricing is $3.50 per 1,000 queries for up to 5,000,000 queries per month:
https://cloud.google.com/vision/pricing
Above that quantity "Contact Google for more information":
https://cloud.google.com/contact/
On Mon, Jun 17, 2019 at 8:23 AM James Forrester jforrester@wikimedia.org wrote:
On Mon, 17 Jun 2019 at 06:28, Yann Forget yannfo@gmail.com wrote:
It has been suggested many times to ask Google for an access to their
API
for searching images, so that we could have a bot tagging copyright violations (no free
access
for automated search). That would the single best improvement in Wikimedia Commons workflow
for
years. And it would benefit all Wikipedia projects, big or small.
Yann,
As you should remember, we asked Google for API access to their reverse image search system, years ago (maybe 2013?). They said that there
isn't
such an API any more (they killed it off in ~2012, I think), and that
they
wouldn't make a custom one for us. The only commercial alternative we
found
at the time would have cost us approximately US$3m a month at upload frequency for Commons then, and when contacted said they wouldn't do
any
discounts for Wikimedia. Obviously, this is far too much for the Foundation's budget (it would be even more now), and an inappropriate
way
to spend donor funds. Providing the service in-house would involve
building
a search index of the entire Internet's (generally non-free) images and media, which would cost a fortune and is totally incompatible with the mission of the movement. This was relayed out to Commons volunteers at
the
time, I'm pretty sure.
Obviously Google might have changed their mind, though it seems
unlikely. I
imagine that Google engineers and product owners don't follow this
list,
so
it's unlikely that they will re-create the API without being asked
directly.
J.
*James D. Forrester* (he/him http://pronoun.is/he or they/themself http://pronoun.is/they/.../themself) Wikimedia Foundation https://wikimediafoundation.org/ _______________________________________________ Wikimedia-l mailing list, guidelines at:
https://meta.wikimedia.org/wiki/Mailing_lists/Guidelines and https://meta.wikimedia.org/wiki/Wikimedia-l
New messages to: Wikimedia-l@lists.wikimedia.org Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/wikimedia-l,
mailto:wikimedia-l-request@lists.wikimedia.org?subject=unsubscribe
Wikimedia-l mailing list, guidelines at: https://meta.wikimedia.org/wiki/Mailing_lists/Guidelines and https://meta.wikimedia.org/wiki/Wikimedia-l New messages to: Wikimedia-l@lists.wikimedia.org Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/wikimedia-l, mailto:wikimedia-l-request@lists.wikimedia.org?subject=unsubscribe
-- Jai Jagat 2020 Grand March Coordination Team https://www.jaijagat2020.org/ +91-74 34 93 33 58 (also WhatsApp) _______________________________________________ Wikimedia-l mailing list, guidelines at: https://meta.wikimedia.org/wiki/Mailing_lists/Guidelines and https://meta.wikimedia.org/wiki/Wikimedia-l New messages to: Wikimedia-l@lists.wikimedia.org Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/wikimedia-l, mailto:wikimedia-l-request@lists.wikimedia.org?subject=unsubscribe