Their examples look much better in some objective sense. Especially if you want to create something that looks realistic in animation.
But I do appreciate the artefacts of StyleGAN2 as an artistic choice, too.
Alias-Free GAN
71–80 of 83 posts
Re: Alias-Free GAN
#72Earlier quoted context omitted.
Thank you!! i spent years trying to understand how it can be that some software that was supposed to corresponds to the well cited paper did not even come close to reproducing the results of the said paper, This was my exact experience. I didn’t understand why I kept having it, and kept blaming myself for not being careful enough. My code must be wrong, or the data, or something . Nah. It was the idea. Kept feeling l…
That seems like a worthwhile thing to publicize in and of itself? The replication crisis in psychology threw out 50% or so of supposed scientific results. If this (or just straight fraud) is common elsewhere, it seems like knowing about that would be a good thing for science.
I agree, most top conferences nowadays publish reviews openly, and address this issue I think. Also it is easier said than done, this is so endemic in so many different academic settings, not just in the US, but also in Europe.
Re: Alias-Free GAN
#73This group of researchers consistently demonstrates a degree of empirical rigor that is unmatched across any other ML lab in industry or academia - remarkable empirical results as always, reproducible experiments, open-source and well-engineered codebase, and valuable insights about low-level learning dynamics and high-level emergent artifacts. Applied ML wouldn't have such a bad rap if more researchers held themselv…
This isn't true. I do ML every day. You are mistaken. I click the website. I search "model". I see two results. Oh no, that means no download link to model. I go to the github. Maybe model download link is there. I see zero code : https://github.com/NVlabs/alias-free-gan Zero code. Zero model. You, and everyone like you, who are gushing with praise and hypnotized by pretty images and a nice-looking pdf, are doing dam…
You won't need that much source code to verify their claims.
Their central improvement is that they limit the generation of high frequencies by ReLU through a upsample-ReLU-filter-downsample sequence.
Their theoretical section explains quite well why high frequencies can be proven mathematically to cause issues. And their practical implementation using filters to cut those off is very straightforward.
If someone tells you "The microphone recording had 50Hz noise so I used a filter to remove it", that's pretty much good enough for someone with experience in the field to replicate their results. This is the equivalent in AI. They uncovered a simple basic issue that everyone else overlooked, but once you know it, it seems obvious in retrospect.
Re: Alias-Free GAN
#74Earlier quoted context omitted.
There is no reason to doubt the open-source release of this work will be any different. Then this is not a scientific contribution yet. We must wait and see. The most important tenet of science, is to doubt. I didn’t even read the name on the paper before I wrote my comment. Yes, I know this group. They’re why I got into ML, along with the group from OpenAI who published GPT-2. Because A+ science. Their claims here a…
Expecting fully executable code to accompany every publication is kind of unique to the modern ML research Scene. As someone from a very different computational research field, where zero code is the norm, not the exception, this reads as a somewhat entitled rant. Reimplementation of a paper is actually a test of the robustness of the results. If you download the code of a previous paper, there may be some assumption…
I think the discrepancy between what actually happens in the code and the main ideas in the paper is a great point, and it touches on the parent commenter's goal of attaining knowledge. For instance, even if the results are reproducible, maybe the proposed key idea in the paper is not the piece of code with the largest impact on performance. Though, with the code available you might be able to discover exactly what the discrepancy is if there is one and you might be stuck at "it doesn't reproduce" otherwise
Re: Alias-Free GAN
#75Earlier quoted context omitted.
> I do ML every day. > I go to the github. Maybe model download link is there. I see zero code Paper was released today. Chill. They said they will release the code in September (I'm guessing late September). The paper is also a pre-print. They're probably aiming for CVPR and don't want to get scooped. > Model first. Code second. Paper third. That's how you produce ML code and documentation but that is not how you re…
> They're probably aiming for CVPR The template screams NeurIPS though. Page limit for that would be 9 pages, this is 9.5, they might have started adding things after the first deadline, anticipating an extra page for camera ready? I mean, that's just a bit of paper astrology of course. But if I'm right, then the author notification is September 28 and camera ready will be due in October, assuming it is accepted. So…
> The template screams NeurIPS though. Page limit for that would be 9 pages, this is 9.5,
That's a fair point and I'm not sure why I didn't consider that they would release a pre-print after they had submitted it. (This is a total fumble on my part)
> there is an issue here about big teams breaking the blind review process by putting themselves on the front
I don't see that as actually breaking the blind review part. There are many more abuses that de-anonymize themselves. Most transformer research is done by big labs because they need the processing power and are the only ones who can afford such equipment (though there was a paper that did transformers on CPUs). Just training ImageNet is out of bounds for a lot of people (I have a few A6000s and it still takes me days). A trivial example is that Google will use JFT and will include it everywhere. If you're qualified to review you're probably going to be able to de-anonymize the lab. I do think we need to do more to make a more level playing field but that's an extremely difficult thing to do. More resources just enables you do do more. But maybe we shouldn't metric hack as much, which would slow things down a little.
Re: Alias-Free GAN
#76Earlier quoted context omitted.
I say chill because you're freaking out. The scientific method is being followed here. Code is not needed for the scientific model to be followed. Even data. Literally every other field is able to advance without public code or data (in fact most areas of CS). There's absolutely no reason to believe that they won't release their code. They have a history of doing so. Models and checkpoints are not the bare-minimum fo…
> If you require someone else's code to reproduce results then you're not convincing me you're a good ML researcher nor programmer. I call bullshit. In computer science, not releasing the code of an algorithm whose output you describe is akin to maliciously obfuscating your methods. No serious paper should be accepted without a script to reproduce the exact same results again.
> In computer science, not releasing the code of an algorithm whose output you describe is akin to maliciously obfuscating your methods.
Well tell that to my advisor (it's also something I've done in the past). So my experience doesn't reflect your claim.
> No serious paper should be accepted without a script to reproduce the exact same results again.
You do realize that this is a pre-print, right? If it went to NeurlIPS then they did release the code to them and will release the code to the public later.
Re: Alias-Free GAN
#77Earlier quoted context omitted.
I have no particular expertise here, but I wonder if you've learned to accept a mostly-broken process? We have the Internet, so why settle for slow diffusion over years instead of rapid communication? Why should graduate students have to spend years trying to reproduce stuff that turns out to be no good? Nobody should have to put up with getting their time wasted like that.
I think it is a social problem, not a technical one. A healthy research field should have some level of cooperation between participants. If you go ahead and publish a "this does not reproduce" paper, you can easily ruin someone's career, so in most cases you don't. I know this is not the platonic ideal of science, but it is the reality, especially of smaller research communities. I agree this is not ideal, but not s…
This concern over ruining someone’s career itself seems like a symptom of a broken process? Making it safe to openly discuss failures is important.
In at least some big companies in the private sector we have “blameless postmortems” where we describe what went wrong in an operational failure without blaming the participating employees.
Re: Alias-Free GAN
#78The previous approaches learned screen-space-textures for different features and a feature mask to compose them. Now it seems to actually learn the topology lines of the human face [0], as 3D artists would learn them [1] when they study anatomy. It also uses quad grids and even places the edge loops and poles in similar places. [0] https://nvlabs-fi-cdn.nvidia.com/_web/alias-free-gan/img/ali... [1] https://i.pinimg.c…
Yes. It's interesting that imposing what are essentially 2d invariance constraints leads the network to learn what we regard as 3D concepts.
Re: Alias-Free GAN
#79That's starting to be high enough quality that you could start considering using that for some Hollywood-grade special effects. That beach morph stuff is pretty impressive. Faces, perhaps not quite there yet because we are so hyper-focused on those biologically, but you could make one heck of a drug trip scene or a Doctor Strange-esque scene with much less effort with some of those techniques, effort perhaps even get…
Re: Alias-Free GAN
#80Earlier quoted context omitted.
I think it is a social problem, not a technical one. A healthy research field should have some level of cooperation between participants. If you go ahead and publish a "this does not reproduce" paper, you can easily ruin someone's career, so in most cases you don't. I know this is not the platonic ideal of science, but it is the reality, especially of smaller research communities. I agree this is not ideal, but not s…
This concern over ruining someone’s career itself seems like a symptom of a broken process? Making it safe to openly discuss failures is important. In at least some big companies in the private sector we have “blameless postmortems” where we describe what went wrong in an operational failure without blaming the participating employees.
Sure having blameless postmortems would be amazing, but I think the informal process I described is probably the closest you will get to it. The reason being that any given subfield can't make this decision in isolation, because the people to who it matters (funding agencies, faculty search committees, journal editors to a degree) are not part of the field, and when they see such a 'blameless postmortem' they will think 'whoa, this person really messed up, we'd better stay away'.
Maybe I am wrong though, and a better culture is possible, like the shift to preprints has happened in a lot of fields and was probably previously unthinkable. So good on you for taking an idealistic stance, I am probably just being grumpy. That being said, whatever culture changes may be beneficial, I stand by my original point that simply dumping code and model alongside the paper is not unambiguously good and may even obscure problems.