Show HN: Deep Learning Shirt for Charity | Hacker News Reader

Show HN: Deep Learning Shirt for Charity

some1else.tumblr.com

10 pointssome1else6 comments

$30 for a shirt with a screen printed image from some random deep learning paper? You don't even cite the paper. Assuming the source is published in some venue then the publisher (ACM, IEEE, etc.) owns the copyright of that image and while i haven't heard of IEEE going after people for copyright infringements this seems to be pretty cut and dry.

Also, using the language, "for Charity", doesn't sound great when you link to a tumblr blog instead of the actual charity website.

The whole thing just looks shady as hell.

The price is steep because charity. I used output from Caffe[1] which is open source software. You can get a similar output yourself, just by running their Python notebook[2]. The festival website is linked within the post[3]. I am using Tumblr to see if it's a viable medium for mini fundraisers.

Thanks for the feedback though, I will set up a landing page instead.

[1]: http://caffe.berkeleyvision.org/

[2]: http://nbviewer.ipython.org/github/BVLC/caffe/blob/master/ex...

[3]: http://kudplac.si/portfolio/tknp/

Looks like the filters from the notebook are a re-organized view of figure 3 from http://www.cs.toronto.edu/~fritz/absps/imagenet.pdf (as the caffe page cites at the top). I'm not a lawyer, so i don't really know if making a 'new' image using the components of the copyrighted image is derivative or not.

> Thanks for the feedback though, I will set up a landing page instead.

Please do, if this is really legit, it's in your best interest to set it up in the most professional way possible. It'd probably be better to get the organizers of the festival to put the content of the post in one of their pages and you link there.

The first layer only contains basic features, which are shared among different sets of images[1]. It seems that the ImageNet photos themselves are indeed very protected[2], so I should probably use another dataset as input. Plugging in my iPhone camera roll should result in very similar filters. It will be very interesting to see how much they differ.

Update: The author of Caffe just got back to me with word that the filters are in public domain, therefore safe to use as design elements.

[1]: http://i.stack.imgur.com/Hl2H6.png

[2]: http://www.princeton.edu/main/administration/legal_complianc...

Doesn't help that the title is misspelled

Hm. I renamed the blog to HELLO WORL.. to make the pun a bit more obvious instead.

Reply on news.ycombinator.com