Social experiment?

Discuss your experiences with and ideas about Stardust@home here.

Moderators: Stardust@home Team, DustMods

mrnaz
Posts: 6
Joined: Sun Sep 17, 2006 5:37 pm

Social experiment?

Post by mrnaz »

I suggested this in an earlier post. I think it is worth at least discussing. The previous thread was closed as my main point had already been discussed in a previous thread. So I started this to see what people think.

Looking at the tracks, it would not be difficult at all to write an image processing algorithm that could at least eliminate trackless slides. From what I've seen so far, a track is easily identifiable by the difference in resolution properties between it and the surrounding features in the aerogel. I'm not an expert in image processing, however I don't think that an algorithm to detect such resolution differences would be difficult to develop at all, certainly not "200 million dollars and deades" as a post in my earlier thread suggested. I would think that any college student doing a final year thesis on image mapping would be able to write one up in a few days if you bought him a pizza and a six pack. Using an algorithm with a very high sensitivity you could at least rule out about, in my estimateion, 2/3 or more of the non-track slides, without risking false-negtives. The rest could then be scanned with differing levels of sensitivity and the resulting features could be tagged and isolated. A single person could then evaluate the isolated images only, easily performing the work of thousands of us star dusters in a day.

There is already software out there that can perform far more complex opteraions on far more subtle images, OCR comes to mind, as does the new breed of automated CAPTCHA solvers. The prime example, however, would be face recognition software, which generally works with far more subtle differences when performing recognition, has poorer images, and has zero ability to go back and verify borderline cases.

NASA doesn't exactly have a small budget. While it is limited, I really don't think they'd have any trouble employing an expert image processing technician for a few weeks' contract.

It is for this reason that I think there is another motive for this experiment, and it has nothing at all to do with stardust. I think that this experiment more likely is social experiment to see just how willing people are to give their time and energy to distributed efforts like this one. In the future, I think, if this is a success, we will see more such public involvement experiments but where actual real work that does require the intervention of a Turing-test passable entity (i.e., a human) to be involved. To me, it is fairly clear that StardustAtHome could be performed by a pre-Turing process, and does not require that many eyes are actually looking at the aerogel.
Mighty Pete
Posts: 146
Joined: Thu Aug 10, 2006 7:47 am
Contact:

Post by Mighty Pete »

One idea.

Is it out of focus or just trackless? To high or too low ? How good is your coding?

The other point being is nobody even knows what a ISP track will look like yet. They have samples but taken at a different resolution and a different contrast.
Nikita
DustMod
Posts: 994
Joined: Wed May 17, 2006 8:33 pm
Location: Indiana, USA

Post by Nikita »

Hey!
Don't talk them into doing anything different! I want to look for the dust! :D
From dust we come
mrnaz
Posts: 6
Joined: Sun Sep 17, 2006 5:37 pm

Post by mrnaz »

Mighty Pete wrote:One idea.

Is it out of focus or just trackless? To high or too low ? How good is your coding?

The other point being is nobody even knows what a ISP track will look like yet. They have samples but taken at a different resolution and a different contrast.
Well I'm not an image processing expert, but I've worked with many before, and I know what they can do.

Focus detection is pretty easy. In fact its a joke. Most digital cameras use image analysis to focus now rather than rangefinders, so the technology is so effective that it can be cheaply built into sub $100 consumer-trash grade gear. Good research-grade systems that don't have to fit into a deveice the size of a mathbox I'm sure could do a far better job, isolating images that are out of focus for re-imaging and accurately gauging the focal point of the surface (more on focal lengths later).

And what do you mean nobody knows what an ISP track will look like? What are you all doing here then? _All_ tracks I've found thus far conform to a very specific set of characteristics, the main one of which is relatively constant resolution over a greater range of focal lengths than the surrounding features. This alone makes them easily isolatable computationally.

Furthermore, what on EARTH is with the lens being at an angle in some slides? Why can't the microscope head be fitted to bi-axial runners and be kept a constant distance from the surface? These devices are CHEAP, they are used in all kinds of industry from drills used in routing to mounting lasers for cutting sheet metal into odd shapes. Such a device would allow them to keep a relatively constant (within a few fractions of a millimeter) distance off the surface of the aerogel as well as eliminating the problem of angular focus.

I see too many problems here that look contrived, which is why I am strongly suggesting that there is more to this experiment than meets the eye. I think that WE are the subjects here, not the dust particles. Everyone, get out your tinfoil hats! :lol:

I do apologose to those of you having fun here (and I do concede that dust hunting is quite addictive hehe), but I'm just engaging in some healthy cynicism. :wink:
Mighty Pete
Posts: 146
Joined: Thu Aug 10, 2006 7:47 am
Contact:

Post by Mighty Pete »

Ya well we are not working in millimetres here we are working in millionths of a inch. And the range finder is having problems determining where the surface is when the surface is no longer perfectly flat. It's all relative though, It's perfectly flat in a millimetre scale just not flat on a micron scale.

I'm sure we are part of the experiment. I'm fine with that. I have the time. I'm always donating computer time to various projects. Things that interest me, things I like to support.
decomite
Posts: 5
Joined: Thu Aug 17, 2006 1:47 pm

Post by decomite »

That's quite exactly what I was saying there.
mrnaz
Posts: 6
Joined: Sun Sep 17, 2006 5:37 pm

Post by mrnaz »

Mighty Pete wrote:Ya well we are not working in millimetres here we are working in millionths of a inch. And the range finder is having problems determining where the surface is when the surface is no longer perfectly flat. It's all relative though, It's perfectly flat in a millimetre scale just not flat on a micron scale.
We are not working in millionths of an inch when it comes to control of the head. Having the head off the surface within tolerance would be easy, and then use digital processors to correct the focus for surface fluctuations. This could be done easily, cheaply and very, very quickly.

Also, go metric!
tim_yoda
Posts: 39
Joined: Fri Aug 04, 2006 5:06 pm
Location: Broadview Heights, OH USA

Post by tim_yoda »

I
see too many problems here that look contrived, which is why I am strongly suggesting that there is more to this experiment than meets the eye. I think that WE are the subjects here, not the dust particles.
mrnaz,
I sent you a PM regarding the "ulterior motives discussion" in past discussion threads. Lo & behold, Decomite lends the same thread in his/her reply in this thread.
In retrospect, I think the only conspiracy with Stardust@Home is to envolve the (global) public and educational sector in space exploration.
It's free.
It's fun.
It's educational and enlightening to read discussions from intellectuals such as yourself and the many other Dusters/Mods/Team.
Let's NOT automate this venture, but expand it other venues.
Tim
bmendez
Stardust@home Team
Stardust@home Team
Posts: 530
Joined: Wed Apr 19, 2006 11:28 am
Location: UC Berkeley Space Sciences Lab
Contact:

Post by bmendez »

Hi All,

Stardust@home is simply a search for interstellar dust collected by the Stardust mission and nothing more. True, that it has wide ranging educational and outreach value (which is actually my primary function), but that is an added benefit and not the primary goal. Finding the dust is the primary goal. It always has been and always will be.

NASA may have many resources, but the amount of funding given for the recovery of the interstellar dust returned by Stardust is relatively small. Using the resources available to us, we did consult with computer scientists here at UC Berkeley. The team here at UC Berkeley has considerable experience with capturing high speed dust particles in aerogel and also pioneered a techique for recovering such particles. We had a great deal of data to present to the computer scientists with whom we consulted. The consensus was that this problem was somewhere between trivial and impossible, and there was no way to tell where along the spectrum it would lie until we actually had the aerogel back in the lab and had started scanning it with microscopes.

We came up with the idea of involving the public in this search after conversations with our colleagues just down the hall from us who operate SETI@home. Instead of asking the public to volunteer their computer's free time to analyze radio telescope data, we are asking the public to volunteer their own time. This is not unprecedented. There have been other projects where humans are more efficient in searching for features in images: Clickworkers, and Spacewatch are two such examples.

We originally planned to conduct both a public search and develop software to search the data in parallel. Even if we could design sophisticated enough software to find the dust, it would need to be taught what the tracks look like. No one has ever captured interstellar dust in this way, so we really did't know what the tracks would look like with enough precision to have the software be successful without at least a dozen examples. So we would need to find most of the dust first before the software could effectively search for the rest.

NASA did not fund our proposal for the full amount requested and we had to scale back the project a bit (this is normal in space science research). We decided to go with the public search because it has a higher probability of successfully finding the dust. The question of speed is not really relevant. If we were starting with all 700,000 fields of view already scanned, had good software and our army of volunteers ready to go, clearly the software would be faster. However, we have to scan the aerogel to create all 700,000 focus movies first. That process is laborious. The automated microscope can scan about 1 tile per day, but for only about 4 days a week (the staff at the interstellar dust lab at Johnson Space Center in Houston have other projects beside this one). There are 130 tiles to scan, so it will take over 7 months to complete the scan. We have enough volunteers that the search will be over only days after the last tile is successfully scanned and uploaded. So, even if computer software were doing the search it would still take a mininum of 7 months to complete. Since people are more likely to find the dust with higher efficiency, that is what we chose to do.

We do encourage anyone out there who thinks they could write software to effectively search for the interstellar dust to do so. We'd be interested in seeing what people come up with. It could potentially be very useful.

Thanks,
-Bryan
"I am made from the dust of the stars, and the oceans flow in my veins"
- RUSH
the moon
Posts: 177
Joined: Thu Aug 03, 2006 12:34 am

Post by the moon »

First of all, I'm not sure what you mean by "tracks have different resolution properties" because they don't. I hope you aren't talking about the calibration movies. Anyway, if you want to know why image processing won't work to find tracks, read this thread. Those are the 2 best possible tracks identified by the researchers, and yet the debate on whether they actually are tracks goes on and on. Truth is, they aren't tracks, the researchers were wrong to even consider them based on those movies. If someone wrote an algorithm to find tracks based on the same assumptions the researchers use they'd get too many false positives to make it worth while.

Software probably could be used to rule out 2/3 of the movies but that still leaves what, 100,000 movies? You'd still need the same system they're using now to view the rest of those, and while you're at all, might as well forget the software and let us view them all. They don't pay us and people enjoy it. Plus the speed and efficiency of the program is only limited by the microscope. It takes forever to scan a cell and to go back and review dust canidates. The vollunteers can review each movie over 100 times so quickly that it doesn't slow down the program at all.

So even if image processing algorithms could be used. What advantages if any would they have over the current system?

As for your social experiment theory, it's obviously ridiculous. I'll humor you and tell you why. NASA isn't interested in that kind of research.
Plus if you're right, that means they aren't actually trying to find intersteller dust, even though it's there in the collector, or is it? Which theory are you actually proposing?
1. The spacecraft did not have 2 sides on the collector, only comit dust was collected and all the slides we see are from some other source.
2. They did collect the dust but for some reason are no longer interested in finding it. They take the pictures from the collector and then ignore our results.
3. They did collect the dust and are secretly using image processing or some other means to find it, while ignoring our results.

As for the angled slides, the aerogel isn't flat, especially near the edges of the cells it's sloped. How does keeping the lens at a constant distance from the surface fix that?

Edit: Doh, I didn't see bmendez's post, so yeah, what he said too. But also in reply to him, you always say that to have software find the tracks, it needs be to first taught examples. But you did that experiment where you shot particules into aerogel on earth at the same speed you expected the IS particules to hit at. So what's wrong with those examples? The tracks might look a little different since they didn't happen in space, but they can't be so different as to be useless.
oscar
Posts: 24
Joined: Tue Aug 08, 2006 3:19 pm
Location: Bellefonte, PA

Post by oscar »

:roll: :roll: :roll:
What we think, we become. Buddha
georingo
Posts: 28
Joined: Sun Aug 13, 2006 2:12 am
Location: Belgium, Gent

Post by georingo »

mrnaz, you've watches to much "Lost"...
Anyway, in the very unlikely event that this is only a social experiment, it would be science too ... and I would be proud to participate :-) !
mrnaz
Posts: 6
Joined: Sun Sep 17, 2006 5:37 pm

Don't take this so personally guys!

Post by mrnaz »

Before I start, may I reiterate that my conjecture is only that: conjecture. Obviously, we'll never really know. But I have a few rebuttals to some of the points mentioned above:

bmendesz: I greatly applaud your getting involved in this, but we can't really expect a confession akin to "Aah mrnaz you caught us! We confess!" now can we? :)
bmendez wrote:So, even if computer software were doing the search it would still take a mininum of 7 months to complete.
That timeframe was the scan time, it is not related to the method of data processing, manual or computational. I.e., it would take that length of time to scan it either way you choose to search for dust.

Furthermore, in the first month of scanning a few tiles I'm sure you would ahve found some examples of what to look for. In the remaining 6 you'd have been able to devise some algorithms to search, so that they were ready by the time scanning completed. There are a number of off-the-shelf algorithms in the public domain that NASA already uses to scan telescope data from Hubble that could be used with a few tweaks and modifications.

I asked a few of my friends in the field and they suggested a book that had some good info on this subject: http://www.amazon.com/Image-Processing- ... 0521599148

In short: I don't really buy the idea that image processing was too hard to develop or would take longer. NASA are already among the foremost experts on it. Just call your pals who work on HST for some pointers and probably some usable code too :)
bmendez wrote:Even if we could design sophisticated enough software to find the dust, it would need to be taught what the tracks look like. No one has ever captured interstellar dust in this way, so we really did't know what the tracks would look like with enough precision to have the software be successful without at least a dozen examples.
I counter that you gave me enough examples to identify tracks, and everyone using SD@H has been given similar examples. So when the plates arrived back you guys must have gone over a few of them searching for such examples to give to us. So that's a staw man argument. I.e., "We will give you examples of what to look for because we need you to find examples of what we are looking for."
the moon wrote:If someone wrote an algorithm to find tracks based on the same assumptions the researchers use they'd get too many false positives to make it worth while.
While there would be false positives, I doubt they'd be so numerous that they'd swamp real data. If we humans select those more than 50% of the time, then that's a false positive error we'd make anyway.
the moon wrote:I'm not sure what you mean by "tracks have different resolution properties" because they don't
They actually do. A track is always almost featureless at the surface, but resolves clearly below the surface. Any "false positives" that cause this are indistinguishable without analyzing the dust particle itself.
the moon wrote:As for your social experiment theory, it's obviously ridiculous. I'll humor you and tell you why. NASA isn't interested in that kind of research.
I can assure you that NASA absolutely most positively IS interested in public interest and involvement in its operations. NASA's budget is pretty much directly related to public interest, given that the politics of the US are such that government "buys" votes by investing in those public projects that are popular. So if NASA can increase public awareness and interaction, it is actually lobbying for more budget dollars.
the moon wrote:Plus if you're right, that means they aren't actually trying to find intersteller dust, even though it's there in the collector, or is it? Which theory are you actually proposing?
None of the silly ones you propose. There is no reason they are not conducting legitimate research on collected dust, using SD@H as a means to generate extra public attention. There is no reason why SD@H isn't entirely non-beneficial either, I am just suggesting that to me, the actual science doesn't seem to be of great enough value to be the primary reason for doing it this way. Whether they are using another method of search as well as SD@H is irrelevant to this discussion.
the moon wrote:As for the angled slides, the aerogel isn't flat, especially near the edges of the cells it's sloped. How does keeping the lens at a constant distance from the surface fix that?
There are pretty mundane devices available in industry to correct for this. Furthermore, are you sure the surface isn't flat? Can we get NASA clarification and perhaps even some photos to verify this suggestion? In any case, even if it's true, as I said there are very simple ways to correct device angle of incidence that have been available for a long time at low cost.

...

Guys, I acknowledge that there is no way this can be proven or disproven. This is an academic exercise I am conducting in discussing this possibility, and I suggest to anyone thinking of posting an angry response that you should remember that you are participating in a scientific project here, and that getting angry because you think something is false is just silly. If there are any logical reasons why what I am proposing cannot be the case I'm all ears!
bmendez
Stardust@home Team
Stardust@home Team
Posts: 530
Joined: Wed Apr 19, 2006 11:28 am
Location: UC Berkeley Space Sciences Lab
Contact:

Re: Don't take this so personally guys!

Post by bmendez »

mrnaz wrote: bmendez: I greatly applaud your getting involved in this, but we can't really expect a confession akin to "Aah mrnaz you caught us! We confess!" now can we? :)
That's the problem with conspiracy theories. No matter how much evidence is presented to counter them, the true believers will always assume that the evidence is false and that their particular theory is correct.
bmendez wrote:So, even if computer software were doing the search it would still take a mininum of 7 months to complete.
That timeframe was the scan time, it is not related to the method of data processing, manual or computational. I.e., it would take that length of time to scan it either way you choose to search for dust.
That was my point, from start to finish this project would take 7 months, no matter which way we did it. I clearly pointed out that an automated system would be very, very much faster. However, it would still finish its work on the same date.
Furthermore, in the first month of scanning a few tiles I'm sure you would ahve found some examples of what to look for. In the remaining 6 you'd have been able to devise some algorithms to search, so that they were ready by the time scanning completed. There are a number of off-the-shelf algorithms in the public domain that NASA already uses to scan telescope data from Hubble that could be used with a few tweaks and modifications.
As I stated, we had proposed to do just that, but we were not fully funded.

I have worked with HST data, and there is no such thing as off-the-shelf algorithims to scan for IS dust particles. And I hope you're not suggesting we should have used IRAF for this work. Yikes!
In short: I don't really buy the idea that image processing was too hard to develop or would take longer. NASA are already among the foremost experts on it. Just call your pals who work on HST for some pointers and probably some usable code too :)
We never suggested that it would take longer, or that it was too hard. In the end, it boils down to "this way is easier."

Some of the most sophisticated pattern recognition software is employed by the likes of Google to filter out "adult" images in searches. We consulted with people familiar with this software who demonstrated just how poorly it does.

Again, Stardust@home, is just plain easier.
bmendez wrote:Even if we could design sophisticated enough software to find the dust, it would need to be taught what the tracks look like. No one has ever captured interstellar dust in this way, so we really did't know what the tracks would look like with enough precision to have the software be successful without at least a dozen examples.
I counter that you gave me enough examples to identify tracks, and everyone using SD@H has been given similar examples. So when the plates arrived back you guys must have gone over a few of them searching for such examples to give to us. So that's a staw man argument. I.e., "We will give you examples of what to look for because we need you to find examples of what we are looking for."
In the tuturial we stated that for the examples of tracks we have used tracks of extraterrestrial particles that were captured in the ODCE collector on the Russian space station Mir, and tracks of submicron dust particles shot into aerogel at 20 km/sec using a Van Der Graaf dust accelerator in Heidelberg, Germany.

We have yet to find an unambiguous example of an interstellar dust track in the aerogel from Stardust. When the first scans arrived we only used them to create blank calibration frames (there were no tracks in them, as that would have been extremely lucky considering the odds against it). All the tracks you see in calibration movies have been digitally cut from the simulated tracks and "dubbed" into the blank calibration movies.

We gave the best examples we could of tracks and warned that actual tracks could be quite different. The most important thing for people to learn from the Tutorial is how to use the system and what not to click on (cracks, scratches, aerogel fragments, bad focus fields, etc.).

People are much easier to teach than software, and are much better at recognizing patterns and being adaptable.
Guys, I acknowledge that there is no way this can be proven or disproven. This is an academic exercise I am conducting in discussing this possibility, and I suggest to anyone thinking of posting an angry response that you should remember that you are participating in a scientific project here, and that getting angry because you think something is false is just silly. If there are any logical reasons why what I am proposing cannot be the case I'm all ears!
I agree that angry posts should be deleted before submitting.

As I stated before, if you are intent on believing that we here at Berkeley had better ways of conducting this research than involving the public and simply are doing this for the publicity or as some kind of social research experiment, then nothing I say will convince you.

All I can say is that I am very happy that Drs. Westphal and Butterworth chose to take this easier approach, because it does allow the public to be involved in a real research project. Everyone has the chance to be a part of ground-breaking research and to see first-hand all the messy details of how it is done. The movies, television, and especially classroom textbooks give the very incorrect view that science is a very ordered and clean process. Science is a human endeavor, and for me, whose ultimate goal is to educate the public about how science is done, there is no better way to show people that then by letting them be a part of it.

-Bryan
"I am made from the dust of the stars, and the oceans flow in my veins"
- RUSH
mrnaz
Posts: 6
Joined: Sun Sep 17, 2006 5:37 pm

Re: Don't take this so personally guys!

Post by mrnaz »

bmendez wrote:All I can say is that I am very happy that Drs. Westphal and Butterworth chose to take this easier approach, because it does allow the public to be involved in a real research project. Everyone has the chance to be a part of ground-breaking research and to see first-hand all the messy details of how it is done. The movies, television, and especially classroom textbooks give the very incorrect view that science is a very ordered and clean process. Science is a human endeavor, and for me, whose ultimate goal is to educate the public about how science is done, there is no better way to show people that then by letting them be a part of it.
All in all, I too think this is a great idea and I hope that more scientific projects involve the public. I often lament the lack of public interest in academic and intellectual endeavor and I hope this and projects like it reverse that trend.

You are right, I will always have a niggling doubt in my mind as to the true nature of this experiment. However, I will not let that stand in the way of me and everyone else here having a lot of fun.

Happy dusting people!
Post Reply