High res movies, more harm then good?

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

Moderators: Stardust@home Team, DustMods

Post Reply
the moon
Posts: 177
Joined: Thu Aug 03, 2006 12:34 am

High res movies, more harm then good?

Post by the moon »

First of all, this is all my speculation since I've never seen a higher resolution movie. It's nice to see that they're shaking things up since the results so far have been mixed at best. But I'm concerned that the new increased magnification movies will only make it harder to differentiate real tracks from inclusions, cracks, and other sub-surface features.

Right now the best tools we have to tell tracks apart from non-tracks are context and focus depth.

By context I mean what else is around the spot in question. For example, other spots at the same focus depth indicate inclusions. Slanted surfaces indicate the spot may be on a lower part of the surface and not actually under the surface. Triangle shapes that are sunken into the surface indicate some kind of contamination and sub-surface features near them are more likely surface debris.

The new movies will narrow your view too much to be able to see many of these important context features along with the feature in question, thereby giving you no choice but to mark a track that you would otherwise be able to rule out. The end result is too many false positives. I know the standard answer to this is, well the best candidates will still get the most hits and rise to the top. In this case, no I don't agree, take a look at this movie
http://stardustathome.ssl.berkeley.edu/ ... =4354669V1
It has the highest hit ratio in my events list and is listed as possible IS particle. But it's an inclusion, you can tell by the 3 similar spots under the surface. Thing is, those 3 spots are all in different quadrants. If you split this movie into 4 parts, you would only see 1 spot at a time, and have no choice but to mark each as a track. Especially the one in the upper right, I believe would be one of the most clicked on high res movies.

Now by focus depth I simply meant that more depth is better. Far too many times I've been unable to make an informed decision on a possible track because the focus cut off too high up. Take that same movie above. Give me 10 more focus bars down and I could say even more definitively that they're inclusions, not tracks. If it was up to me to take this project into a phase 2, I would have dealt with the focus depth issue, not the resolution issue.

I'm convinced that the resolution was already sufficient to clearly make out any real tracks. But please prove me wrong!!!! Can we please please please please see examples of some of the better track candidates you're planning to extract? I think that would clear up so many issues for so many people.

fflo
Posts: 63
Joined: Tue Aug 08, 2006 7:37 am
Location: Eindhoven, Netherlands

Re: High res movies, more harm then good?

Post by fflo »

the moon wrote:First of all, this is all my speculation ....................................................................In this case, no I don't agree, take a look at this movie
http://stardustathome.ssl.berkeley.edu/ ... =4354669V1
It has the highest hit ratio in my events list and is listed as possible IS particle. But it's an inclusion, you can tell by the 3 similar spots under the surface. Thing is, those 3 spots are all in different quadrants. ......................................................................................................................................I'm convinced that the resolution was already sufficient to clearly make out any real tracks. But please prove me wrong!!!! Can we please please please please ..............................................
O.k. then, I am not sure that I can prove you wrong, but viewing your above movie I must make the following observations:
With my (Netscape) browser, I can clearly detect two sub-surface features. I am not able to determine, whether they are inclusions or not.
Since you can only click one as possible track, I suppose that in your events list the number of hits is distributed over these two features.
Now, when I should get this movie divided into four, I would have the chance to click both possible tracks and so highten
a) the number of hits for both of these "candidates" and
b) their chanes for further examination.
Furthermore, I don't think that more focus bars would avoid "bad focus" movies. But those we can mark for further examination!

Maybe your speculation got to be revised -- a little :?:

In the meantime:
From dust we come -- for dust we search -- to dust we go

fflo

jsmaje
Posts: 616
Joined: Tue Aug 15, 2006 8:39 am
Location: Manchester UK

Post by jsmaje »

Sorry fflo but I have to agree with the moon's every word (by the way, the 3rd probable inclusion is just coming into focus about an inch to the right of the top left corner).

When I asked here a month ago about the new hi-res movies, Bryan seemed to dismiss my suspicion that the project was in effect starting over again. Yet we learn from his latest Phase Two Update that "The scores, rankings and candidate lists from Phase 1 will be frozen, preserved, and displayed on the Website. We will give all the volunteers a fresh start and invite new volunteers to continue to join the search. For Phase 2, everyone will begin with a clean slate for score and ranking and an equal chance at discovering yet more candidate stardust tracks."

As I said, this is science and I'm happy to continue, though others may now feel that their low-res efforts and scores so far are to be considered of little consequence. And if anything is clear it's that with a third of the collector examined nothing has been found so far that convincingly matches the predicted IS tracks. So I'm not sure that the new hi-res calibrations Bryan mentions, based on as yet unproven features, will necessarily be of much value.

Nevertheless, I'm already bored with Galaxy Zoo - Stardust is much more fun.
John
Last edited by jsmaje on Wed Jul 25, 2007 1:01 pm, edited 1 time in total.

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

Post by Nikita »

One thing to consider is that the "old" movies had hits from the very beginning, when we were all still learning. A frequently viewed movie with inclusions is going to have a lot of marks on it. This is our chance to see the items clearer. Better or worse? Well, I don't think it can hurt. Besides, perhaps we can look at the new ones and be able to tell without a doubt what they are. We will see when it happens. I think it is a start over, but better.

It is important to me to send congrats to all the searchers from the "old" search for an excellent job and a good luck to everyone starting again and starting for the first time.

Don't forget to continue to check on the forum as we wait for the project to come on line again, it's a great time to relax and have a little more fun! Also, perhaps time to get to the eye doctor and get ready for the next level.
From dust we come

the moon
Posts: 177
Joined: Thu Aug 03, 2006 12:34 am

Post by the moon »

I thought about this more. Since there's really nothing that can be done about the shallow focus depths without rescanning tiles, I suppose the team is doing the best they can with the data they already have. Even when they do get more microscope time, I'd rather they finish scanning the rest of the tiles instead of getting better looks at old movies.

That being said, I still don't like the direction they're heading. What we need is actual higher resolution views at the same magnification level we're seeing now. Not the same resolution in a zoomed in area. The new high res movies will be the same file size as the old ones right? Just covering a smaller area.

Instead of spliting the high res movies into 4 parts, why not just show us all 4? And make the image bigger, its 500x375 now, is there some reason it can't be 800x600?

If bandwidth is an issue, just slow us down somehow, theres lots of ways, limit the number of movies you can see each day, cap the download rate, etc. Quality shouldn't be sacrificed at all for the users' convinience. It's clear there's more then enough volunteers and we have all the time in the world since any actual work in the lab takes infinitely longer.

greuti
Posts: 200
Joined: Fri Aug 11, 2006 10:11 am
Location: Switzerland

Post by greuti »

We should get a better view on the scans. The new movies (JPEGs) are reduced in file size/quality too but starting from more details (magnification) than the old ones. So a tiny particle/track will be visible at more pixel in the VM (on your screen) than before - therefore more resolution too. Even it could be that the track was too small for appearing as a dot on the screen.

You can already have a look at new movies if those here are already the new ones.

den@s
Posts: 2
Joined: Mon Jun 18, 2007 3:29 am
Location: Uk

Post by den@s »

Having examined your pictures would not have agreed particle present. I also look for ridges on the depression around the ?particle I am hoping that with Better deffinition we will be able to see these rudges as the item penetrates the substrate more clearly All penetrations have to leave marks which will be uniform with depth. 8)

Post Reply