9 hours to figure out the 3 dimensional rotation of the chair?? Wow, I don't think that solution is the path to go down. Why not allow the user to put down some sort of control points to narrow down the search space? You know, it's okay to ask the user for help sometimes.
*Edit: The reason it took 9 hours is because they rendered the model in every single possible camera position and orientation until they found one that matched.
Another problem,is that they need to have a 3D model ready of whatever it is you're trying to manipulate. Have a slightly different chair or different car make? Tough cookies.
What is wrong with researchers getting money directly from the people that are interested in this thing?
You'd rather the government act as a mediary and psychic as to what the public wants? So that they can then apportion the public "research" spending budget accordingly?
Give your money to whoever you want; scientists, charities, ISIS...
> government act as a mediary and psychic as to what the public wants
A lot of that funding allocation is vetted by other academics, as in professors and researchers who are experts in their field (e.g. NSF has a panel of said experts that take time out of their schedule to go to D.C. to read and discuss proposals).
Yeah, and it doesn't look like you get anything for your money, either? It wouldn't be very nice if I paid $20 and helped them make a product only to have to buy it later, or if nothing but a research paper came out.
This looks more like a donation than crowdfunding.
There is already a fair crowdfunding scheme for science called "taxes".
I feel the increase of popularity of crowdfunding schemes for science or health care is detrimental to the cause. They will never be able to approach the same level of funding that a state can give away, yet they will be a good excuse for the politics to reduce the amount of money spent on these issues: "The crowdfunding schemes are already taking care of it, aren't they? Why should we waste taxpayers' money? If prof. X wants its experiment to be funded why doesn't it ask for money on indiegogo?"
For those who may thing "nobody is going to ever suggest that", well, the current government of UK as been elected exactly on this promise: http://en.wikipedia.org/wiki/Big_Society
You should consider the proposition that funding science with taxes is a great excuse to politicize science. Moreover, the notion that science funding could ever be objectively "fair" is complete hokum, and the idea that a political apparatus could find a particularly fair solution is even crazier.
Of course, I have to disclaim that I ran a successful science crowdfunding campaign (http://pledge.indysci.org). As to why it couldn't be funded by taxes, it's because the NIH coordinator for the subsection working in the division that would have funded it used to lead the project I'm continuing so there's a conflict of interest.
Now, this is a relatively 'boring' and particularly coincidental reason for the inability for taxpayers to be able to fund this project. Usually the real reasons are internecine conflicts within sciences competing for funds, personal differences, political connections, etc. Consider for example the case of Leo Paquette, who denied funding to a group and then turned around and stole the idea for himself. (ScienceWeek. 1998-03-20) Of course the system came to a settlement whereby Paquette got a slap on the wrist (agreed to not accept funding for 2 years) instead of kicking him out.
My personal opinion is that if people want to say they fXing love science, then they should be willing to reach into their wallets and pay for it themselves. Science has a clear social benefit and to that end, everyone should want to help out. Forcing other people to pay to make you feel like you have chipped in yourself - is just patting yourself on your own back for making zero personal sacrifice.
There's also already a crowdfunding scheme called "the private market". So far its existence has not caused governments to stop funding science.
The way I see it is that each of these schemes allows money to be allocated according to different sets of criteria, each of which we hope will overlap with the vaguely understood criterion "benefits society". Since the overlap isn't perfect (sometimes pathologically so), I'd argue that diversity of crowdfunding schemes is a good thing.
Crowdfunding science asks the 'Average Joe' to decide what's important in research, and as i shouldn't have to explain, research outcomes aren't as cut and dry as when you fund a video game or movie.
There are more than a few traditional means of obtaining money toward science. These methods aren't perfect, but are reviewed by a panel of peers in their field.
A professor asking for money on indiegogo circumvents this process, which acts as a validation of the authenticity and legitimacy of said research.
A student doing so is even worse in my opinion. Students are funded through the grants or independent scholarships. Their research tools are the responsibility of their advisor or the department. In this case the school and professor are well funded, so asking for 12k (which is about half of a single semester of tuition) is somewhat shoddy.
> Crowdfunding science asks the 'Average Joe' to decide what's important in research
Well, why should the average Joe not be enabled to say "hey, I'd like to take some of my money and put it towards a research project that interests me"?
>and as i shouldn't have to explain, research outcomes aren't as cut and dry as when you fund a video game or movie.
I'm not sure what the criticism is here.
>A professor asking for money on indiegogo circumvents this process, which acts as a validation of the authenticity and legitimacy of said research.
All it does is act as a pre-filter. It's not like funding guarantees publication in a peer-reviewed journal. And that filter also blocks out a lot of potentially useful research.
I'm guessing that filter is very useful for large, expensive research proposals. But advances in technology have made it so that far more research can be done at a much lower cost. With greatly reduced risk, the need for such a filter is also greatly reduced.
> Well, why should the average Joe not be enabled to say "hey, I'd like to take some of my money and put it towards a research project that interests me"?
No problems, like i said in another reply, give your money to whoever you want. What bothers me is them asking.
> >and as i shouldn't have to explain, research outcomes aren't as cut and dry as when you fund a video game or movie.
I'm not sure what the criticism is here.
I do not look forward to the day where students are required (as in to be competitive) to build an a product/hand out t-shirts/etc for the 'Average Joe' in order to get the tools they need to complete their degree.
> All it does is act as a pre-filter. It's not like funding guarantees publication in a peer-reviewed journal. And that filter also blocks out a lot of potentially useful research.
Well, yea, of course. The process has it's own problems (promising too much and not delivering, using buzz-words to be trendy, etc) because it's competitive, and competition breeds all kinds of tricks to gain an edge. I'm worried/uneasy that the same tricks/edge will be used in crowdfunding with larger results.
How many failed crowdsourcing articles are there despite successful funding (e.g. Clang)? Research often fails, but for different reasons (method is currently impractical, no longer necessary, and/or didn't live up to the hype), and putting a spotlight on what didn't deliver doesn't mean the approach should no longer be funded/trusted/used.
In addition, a long term fear is that if the public sees research getting a lot of money outside of traditional channels maybe there'll be a debate as to whether to continue funding through traditional channels. Which brings me back to printing t-shirts so we can advance the state of the art.
Unless I'm missing something, for this to be useful, at least to normal people, there would seem to be a huge step missing whereby every object in the world has to be modelled and put in a database. Also you'd have to be able to locate your 2d objects in the 3d database, but that is at least theoretically tractable I guess.
Edit: It sounds cool, but maybe a bit misleadingly advertised.
I think that 3D scanners would be a perfect solution to this. It becomes easier and easier to digitize real world objects, even based on a single-camera video.
That stuff looks cool, and I want to play with photogrammetry stuff when I have some spare time. I think it's a bit optimistic to say it's a "solution" to this though. It's another tool that can help with the modelling + positioning + texturing + lighting + rendering + compositing workflow, but I don't think we're anywhere near a "3D photo editing tool" that can be used by normal people.
Cool application.
I don't see much justification for double precision arithmetic in this case though, so why would they need the Teslas? Going with Titan Z cards instead I believe they'd come out with comparable single precision performance in the ballpark of half the price? That might be closer to what could be reasonably expected in the target workstation too...
I'm just curious about her degrees. It seems she achieved both a B.S. and M.S. at the same time from the same university. How is that possible/allowed? The schools around me don't even accept people into their masters programs until after you have completed a B.S. and have some relevant work experience as well.
No but really, I think it's more beneficial for a student to take a year or two off after completing their BS and work in their field. The experience you pick up in the workplace actually doing the thing you went to school for is invaluable. That is unless you worked in your field prior-to or while you were attending university.
Some universities, such as the one I went to, have degrees that combine some normal B.Sc degrees with honors. Such that after completing it, you can start with your masters.
Rochester Institute of Technology (RIT), the school in question, has a dual degree program where you can use some graduate classes to satisfy both the BS and MS credit requirements. She still had to write a research thesis.
This is amazing technology, I'm excited for it to actually come to photo editing/manipulation. Sometimes I wish the programming world and academic world intersected more to solve problems.
You can already do it. Use content aware fill to cover your object, then find a 3D model of your object and composite it over the top. Afaict this is intended to streamline that process a bit, but it's basically something you can already do, eg using Gimp + Resynthesizer and Blender.
I don't believe 90% of the examples in that video. They perfectly fill the gaps left by the object every single time, including the person's fingerprint, the carpet pattern, the patterns on the road.
"""
We compute a mask for the object pixels, and use this mask
to inpaint the background using the PatchMatch algorithm [Barnes
et al. 2009]. For complex backgrounds, the user may touch up the
background image after inpainting.
"""
So they're not really claiming perfect object removal is part of their thing. What they are actually doing is quite interesting, but their pitch is making some rather extravagant claims IMO.
*Edit: The reason it took 9 hours is because they rendered the model in every single possible camera position and orientation until they found one that matched.