Dear “Prof” James,
1: We literally state: “As evinced by the papers we cited, privacy preserving obfuscation of images is neither a novel idea and most certainly not our idea”.
Also, Google’s “Inclusive Images Competition” announcement was made on September 6, 2018 (that attracted 468 teams) whose FAQs verbatim included the following questions:
Q. Why are faces in the Challenge data set blurred?
Q. At what stage in the collection and donation process was facial blurring applied?
Q. Won’t the blurring of faces impact the quality of predictions?
Q. Are competitors allowed to try and attempt methods that are targeted towards unblurring the blurred faces?
Q. What if someone tries to create a Person detector by looking for blurred portions of an image and guessing there’s a person there?
2: You mention: “I wouldn't cite that ImageNet Roulette project as IMO it is just a PR stunt given that one of the co-authors was also quite aware of the NSF proposal and in fact a collaborator on the project.”
We weren’t aware of the PR-stunt/NOT-PR-stunt filter that academic disseminations ought to process relevant literature through before citing the works. This framing, in our humble opinion, is a disingenuous political move that we’d rather not advocate or engage in. But glad to know you indulge in these citations antics as a ‘professor’ at Stanford.
3: With regards to the NSF grant proposal you cite, are we talking about this one?
NSF grant#1763642. Link: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1763642&HistoricalAwards=false
We read the grant abstract. Can you please let us know what it is that went amiss in our monologue?
4: At the risk of sounding like a broken record, the main issue we were critiquing is the culture of large scale datasets. You are free to parse through the slide deck dated October 21, 2012 (1530:1600) and head to slide-24 titled: “Ladles are hard” here.
http://www.image-net.org/static_files/files/ilsvrc2012.pdf
So the curators knew of this issue in October 2012 but got the NSF grant process going in 2018.
This means that it took 6 years for the curators to address ethical transgression in the dataset (A simple Google scholar search will reveal an entire body of ethics literature entailing dozens of highly cited papers that talk about consent in digital photography way before 2012 BTW) this is the kind of culture that compute vision is situated in (one refuses to take responsibility) and the culture that we are trying to advocate we move away from.
5: “Timelines do not prove someone's intent”.- Awesome! Thanks for piece of wisdom.
Will put that on a t-shirt :)
