apu

11 ms·

61.
by apu 12y ago
No, the 97.53 is more fair. See here: https://news.ycombinator.com/item?id=7638269
62.
by apu 12y ago
The problem with more pairings is that you have a greater chance of identifying the right person because of their additional photos, but that gets offset by the huge number of photos of everyone else in the database that you will now al
63.
by apu 12y ago
I'm not sure that's the distinction I'd use. To me, it seems more to be a matter of the kind and amount of training data you have. If you have a ton of labeled data (either for the exact task you're doing, or a related o
64.
by apu 12y ago
People are really good at face recognition in real-life, because you have not just a static 2-d view of a person completely out-of-context, but a fully dynamic 3-d view of someone you've probably seen before and you know the lighting
65.
by apu 12y ago
Sure, but the numbers in the research paper are "pure" (i.e., not using all this additional information). In production, I'm sure they must be using all these additional cues as well.
66.
by apu 12y ago
No. Alignment is done by every face verification method, and even the "new" 3d alignment done here is on par with other alignment methods in practice today. At their talk at CVPR, they made clear that the biggest factor in getting
67.
by apu 12y ago
Here's the paper: https://www.facebook.com/publications/546316888800776/ The methods are quite different, but I think at root, it's just more being able to take advantage of much more training data. The
68.
by apu 12y ago
sigh no, it doesn't "outperform humans" on "face recognition". In particular, see my previous comments [1] and [2] for discussion on why this method might be doing well. As for "outperforming humans", a
69.
by apu 12y ago
Where is this from?
70.
by apu 12y ago
I can't tell if you understand parent's comment and are playing along or genuinely missed the point...
71.
by apu 12y ago
Heh, a nice piece of post-hoc rationalization. Compare to this article in the newyorker from 2 years ago (different author, but still...): http://www.newyorker.com/reporting/2012/05/14/120514fa_fact_...
72.
by apu 12y ago
It's black-on-white for me.
73.
by apu 12y ago
This is exactly the sense I get. Plus, he's a programmer ("one of us") and a lawyer ("one of them") and can thus bridge the gap between the tech world and the "real" world -- if anything, we should be hopi
74.
by apu 12y ago
But it's often not clear what the outcome of research will be -- unexpected results can start off as being simply the next iteration of an existing technology.
75.
by apu 12y ago
It's a really phenomenal film. It was shown a couple of years ago at SIGGRAPH's annual animation festival with John Lasseter (of Pixar) introducing it and saying it was one of his favorite films and a real inspiration to him. Fred
76.
by apu 12y ago
These are also related to "Integral Images" or "Summed Area Tables" [1] which are used in computer vision and graphics (resp.) for many purposes (most prominently in the Viola-Jones face detector). The benefit is that ta
77.
by apu 12y ago
The other responses to your question tell only part of the story. Yes, the academic groups that work in deep learning publish papers describing their methods. But these papers are rarely sufficient to be able to recreate the models they bui
78.
Don’t Blame Big Cable. It’s Local Governments That Choke Broadband Competition(wired.com)
162 points by apu 12y ago|152 comments
79.
by apu 12y ago
Although not widely known, the best way to learn about specific computer vision topics in detail is usually through the websites for tutorials held at the premiere computer vision conferences. These are CVPR, ICCV, and ECCV. For example, he
80.
by apu 12y ago
While you make a valid point (for some people), please try to be more civil. Remember that a real person spent time & energy working on writing and releasing this app. You likely wouldn't say your last sentence to this person in re
81.
by apu 12y ago
This kind of comment is not appreciated here.
82.
by apu 12y ago
Also note how cavalierly and routinely the Times (and most other "traditional" publishers) avoid actually linking to this thread, or basically anything other than internal links to other Times articles.
83.
by apu 12y ago
Generally, new results aren't added to LFW until after the conference takes place. CVPR will happen at the end of June this year.
84.
by apu 12y ago
In some sense, the 97% is indeed the more fair number to compare against, assuming that this paper also restricts the algorithms to only see tight crops of the face. Obviously using more than just tight crops will give you more information,
85.
by apu 12y ago
The goal is to allow for "fair" comparisons. But I agree with you in general. I suspect even LFW's creators might; I don't know if they expected the benchmark to still be tested on this many years after creation. I'
86.
by apu 12y ago
First, my earlier comments on a "competing" approach from Facebook may help give relevant context for how to think about these numbers: https://news.ycombinator.com/item?id=7393378 Briefly skimming through this pa
87.
by apu 12y ago
I want to chime in and say that I've found your blog wonderful, and since you're too modest to post a link, let me do it: http://jeremykun.com/
88.
by apu 12y ago
Lytro is also "just" simulating it, but with slightly more/different data. They capture a light field but that doesn't magically give them depth values; they have to estimate them using an optimization, and then render t
89.
The problems of HFT, Joe Stiglitz edition(blogs.reuters.com)
5 points by apu 12y ago|0 comments
90.
by apu 12y ago
the Wall Street Journal has been reduced to reprinting tweets as news Why 2024? I've already seen this happening for the last few years, where the established press takes original content/reporting/etc. from online sources

More ›