calibration movie thoughts (again!)

Archived here are older posts which are no longer relevant or were redundant.

Moderators: Stardust@home Team, DustMods

entropydave
Posts: 15
Joined: Wed Aug 02, 2006 11:28 pm

calibration movie thoughts (again!)

Post by entropydave »

Hi,
I am really amazed that some people have, say, 400 calibration movies correct and only 7 incorrect - I am a reasonably experienced microscopist (did microbiology at Uni - have also used SEMs, petrological scopes and also ordinary electron microscopes as well as telescopic hardware) and consider myself reasonably observant with this tool, yet I can assure you I have not scored anything like the pretty much 100% perfect results some have been reporting... My work years ago in a path lab looking at smear tests and such like have always been pretty much 100%.

I've never had anyone say over the past 25 years that I was not very good with an eyepiece with respect to my observational skills - perhaps they have all been too polite! - but obviously I'm not as superhuman as some of the bods on this project as I have 431 correct to 114 incorrect.
I do find this a little discouraging I must say, so I have now taken to ignoring my scores and just hitting away to the best of my ability at he slides!
"I have the most excellent proof that x^3=y^3+z^3, however my signature is too small to contain it..."

peterm9911
Posts: 4
Joined: Wed Aug 09, 2006 4:36 pm

Calibration movies

Post by peterm9911 »

You should be able to spot the cal movies with tracks, they are all distictly "solarised", usually the tracks are pretty obvious but if they're not you can still see that it's a cal movie with track from the look of the thing. But I'm trying to pretend it's not that obvious and actually treat it like a real movie, even clicking "no track" if I can't see it quickly, despite knowing there is a track in there.

icebike

Re: calibration movie thoughts (again!)

Post by icebike »

entropydave wrote: Hi,
I am really amazed that some people have, say, 400 calibration movies correct and only 7 incorrect - I am a reasonably experienced microscopist
But I bet your percentages are both pretty high right?

I found that my EARLY calibration movie errors are usually imaginary tracks (false positives), and my more recient ones are missed tracks (false negatives). I virtually never get a false positive anymore.

There are a few calibration movies with very small tracks which I missed
till i lowered my screen resolution to 1024x768.

Also some people pay extreme attention to the ondeck_movie_Type and are thereby alerted to a testing movie. I ignore that now, In fact I've turned off my status (inorder to fit in 1024x768, and I therefore don't see it at all.

Forces me to do a good job.

icebike

Re: Calibration movies

Post by icebike »

peterm9911 wrote:You should be able to spot the cal movies with tracks, they are all distictly "solarised"
Not helpfull.

The idea is to ALLOW the CMs to train your brain, not to out-fox them.

The object is not to learn how to spot CMs, but rather to spot tracks.

dustmin
Posts: 4
Joined: Fri Aug 04, 2006 9:32 pm

Re: calibration movie thoughts (again!)

Post by dustmin »

icebike wrote: ..some people pay extreme attention to the ondeck_movie_Type and are thereby alerted to a testing movie. ...
Unfortunately, I think that that was very true.
However, it looks like they have modified the system so that this is no longer obvious in the URL or status bar.
It is a good idea. Some people were just rushing through any "real" movies to just get to a calibration one to up their rankings.
Of course, now I have already have heard from one person complaining that it is unfair that they have changed the labeling as those that already had obscenely high "scores", will probably be locked into a high position as others can no longer "cheat". :? Sigh. Too bad that some seem to believe that this is some sort of a game or contest, rather than data analysis; but I guess it is human nature.

Nikita
DustMod
Posts: 994
Joined: Wed May 17, 2006 8:33 pm
Location: Indiana, USA

Re: calibration movie thoughts (again!)

Post by Nikita »

entropydave wrote:Hi,
I've never had anyone say over the past 25 years that I was not very good with an eyepiece with respect to my observational skills - perhaps they have all been too polite! - but obviously I'm not as superhuman as some of the bods on this project as I have 431 correct to 114 incorrect.
I do find this a little discouraging I must say, so I have now taken to ignoring my scores and just hitting away to the best of my ability at he slides!
Hi entropydave!

I think you made the best choice! Just do your best and have fun!

Good Luck!
From dust we come

Driven1
Posts: 20
Joined: Sat Aug 05, 2006 7:27 am

Post by Driven1 »

I do have a high Ranking at the moment. I could care less about the scores. The Rank says how you're doing in the accuracy department, which is what I feel is most important.

I have never looked at the URL to find out if a movie was real or a fake. Yes, it is very easy to tell the "Doctored" calibration movies from the real ones.

I have treated every movie the same and have been pretty darned accurate. Both my Sensitivity and Specifity are currently running at 99.70.

Very recently, they have added a lot of real movies that have been verified as having a track or no track and have also added some "doctored" movies that have had no tracks implanted in them. I could swear the other night that they were purposely thowing them all at me at once! The only ones I missed had to deal with possible "lateral" tracks. Had I not missed those my Rank would be even higher.

The point is that it has become much harder from a calibration movie standpoint and there are those that WILL get burned on a lot of them. Some of them really got me thinking. In the end, you make the best decision you can and go with it. They'll statistically filter it out later.

The biggest part of this is knowing what you are looking for. You will find most possible tracks very near, or at the very end of the focus range. In most cases they will look similar to the ones in the calibration movies. Some large, (although I haven't really seen any that large as of yet) and some very. very small. This is where I start, at the bottom of the focus range. I look for a possible track and then I focus up to see if there's enough focus range to work with. If there is, and I see something at the end of the focus range I mark it. If there's poor focus period, isn't enough focus range, or I don't see any tracks, I move on. It's actually a very fast way to do it. Aftere a while it's almost automatic.

If I run into something I'm not so sure about, like one of those movies that looks like it has a thousand grains of sand on it, I move my cursor rapidly up and down the focus range and watch different areas of the movie as I'm doing it. This will verify if there's something there or not. If I do spot something (or think I have) I do the same thing to verify it. I then mark it or dismiss it. And yes, there are some that I've had to think about, or flipped a mental coin on.

Edit...

I thought of something else last night that I should have added to this so I'm adding it now. Computer Screen Resolution.

I used to work as a Computer Tech Support Technician for a major Broadband ISP. This is an area I know like the back of my hand.

A screen resolution of 1024x768, true color, is best to view the movies. If you don't know what your screen resolution is, you can check it and change it if necessary. I would expect a size of 800x600 is also acceptable and that a resolution of High Color would also be fine. I wouldn't go any lower than that. Even though these are black and white photos, the color range and size has a lot to do with being able to focus and see tracks accurately and easily.

To check your screen resolution, right click on an empty spot, on your desktop, left click on "Properties" within the menu that appears. Click on the "Settings" tab. This is where you'll find your Screen Size and resolution settings. Note that if you have an older computer, these settings may not be available.

Another quick note related to computers. Those people here with really high Scores are using High Speed Broadband Connections. Those using Dial-up can not ever expect to attain a high Score. The best dial-up users can hope for is to keep their "Specificity" and "Sensitivity" stats up. These are the most important anyway.

Don't sweat your stats! It'll all come out in the wash later. What's important is that you're here and doing the best you can!
Last edited by Driven1 on Thu Aug 10, 2006 7:30 am, edited 1 time in total.

Nikita
DustMod
Posts: 994
Joined: Wed May 17, 2006 8:33 pm
Location: Indiana, USA

Post by Nikita »

Yeah, what he said!
From dust we come

peterm9911
Posts: 4
Joined: Wed Aug 09, 2006 4:36 pm

Re: Calibration movies

Post by peterm9911 »

It not that I tried "to learn how to spot them", it's just that they are pretty obvious. Not my fault if they look so different...... :-)

icebike wrote:
peterm9911 wrote:You should be able to spot the cal movies with tracks, they are all distictly "solarised"
Not helpfull.

The idea is to ALLOW the CMs to train your brain, not to out-fox them.

The object is not to learn how to spot CMs, but rather to spot tracks.

ToSeek
Posts: 48
Joined: Thu Aug 03, 2006 8:40 am
Contact:

Re: Calibration movies

Post by ToSeek »

icebike wrote:
peterm9911 wrote:You should be able to spot the cal movies with tracks, they are all distictly "solarised"
Not helpfull.

The idea is to ALLOW the CMs to train your brain, not to out-fox them.

The object is not to learn how to spot CMs, but rather to spot tracks.
Yes, I'm feeling that I'm being trained more to recognize low-resolution JPEGs than to recognize tracks. They should go back and try to do some higher-quality calibration movies.
If you're going to be just like everyone else, what's the point in existing?

geckzilla
Posts: 30
Joined: Mon Aug 07, 2006 11:05 am

Re: Calibration movies

Post by geckzilla »

icebike wrote:
peterm9911 wrote:You should be able to spot the cal movies with tracks, they are all distictly "solarised"
Not helpfull.

The idea is to ALLOW the CMs to train your brain, not to out-fox them.

The object is not to learn how to spot CMs, but rather to spot tracks.
Oddly enough, I found that before I even realized that the majority of CMs contained significant banding across gradations I had become conditioned to responding "There's something there..." before the movie had loaded and before I had even scanned. It took about 500 CMs for me to become conscious of this. The banding conditioned me into "knowing" it was a CM. Good old Pavlov and his dog.

It's also worth noting that I had yet to discover the movie type being revealed in the URL at that point. Even after I realized that I also found it easier to ignore the URL completely.

peterm9911
Posts: 4
Joined: Wed Aug 09, 2006 4:36 pm

Cal movies

Post by peterm9911 »

I haven't even looked at the URLs :-)

Sharqua
DustMod
Posts: 245
Joined: Thu May 18, 2006 1:02 am
Location: Bradenton, FL

Post by Sharqua »

I'm pretty close to that figure, but I'm closing in on 700 movies with 7 incorrect (at least a couple of which were incorrectly marked movie errors).

I actually close the address bar so I can't see the URL. (I hit the F11 key in Internet Explorer)

I admit I've gotten to where the calibration movies jump out at me, but I'd expect that considering the quality variance. There are a number of calibration movies, however, where you can't detect the lower quality of the JPG. So I must argue that to capture a higher accuracy rating, you need to actually be looking for something in the movies themselves.

Personally, I locate the surface with my mouse, then I slowly move the mouse up and down the bar below the surface point until my eyes are satisfied that the focus of everything in the image is changing normally. Particles on the surface tend to move out of focus in a steady and consistent manner -- and when I say consistent I mean everything moves in an equal fashion.

Working with astronomy images tends to train your eye to look for unusual motion. Looking through motionless images (ie, moving down one focus bar at a time and focusing on each bar one at a time) doesn't work. The motion itself, plus focusing on one quadrant of the image at a time, allows your eye to discern objects that aren't moving as you'd expect.

I must wonder if the people whose accuracy ratings are high have prior experience in flickering astronomy images, be it Spacewatch, Sungrazers, or what have you? I've done both of those, and I found the Sungrazer experience to be the most helpful.

Mind you, I'm only an amateur astronomer at best, and probably not even that, but stuff like this can be done by the layperson at home during the daytime (not Spacewatch -- their funding was cancelled) in the city.

Life is good.

-Shar

bghind
Posts: 2
Joined: Wed Aug 09, 2006 11:56 pm

Calibration Suggestion

Post by bghind »

I would like to suggest that those of us with more than 500 calibration movies and over 97% accuracy should not have to continue to look at calibrations. It is getting so obvious now that I can spot where the dust particle is going to be before I start to focus the movie!!! I've also noticed that you have to have a long down focus to see some of these dust particles (calibration) so I've started to click "Bad Focus" for those movies that don't have at least half the focus field below the surface.
Bill G Hinds

mdiehn
Posts: 1
Joined: Wed Aug 09, 2006 5:56 pm

The Intuition is fun...

Post by mdiehn »

I've only done a bit of this - just a few hundred movies so far. I noticed pretty soon that when I get a "hmmm" feeling about a movie, I should stick with it a bit and just... uhm... doodle, really. Sometimes the feeling is really vague, like, I'll notice that I've been fiddling with the focus for a while or that I'm just sorta staring. Sometimes I just *know* immediately and it takes only a second or less to "confirm" my intuition. Other times, it takes a while, but eventually, something pops out at me. So far, my accuracy is good 100/97 over 414. Total Real Movies is 296.

So, I just started yesterday and I'm going slowly and taking my time. The most facinating part for me is "watching" my intuition incorporate new cues.

Post Reply