Phase 3 calibration - am I wrong?

[heyjude]

Hi.. I have gotten a couple of results I didn't really expect on calibration movies for Phase 3. I didn't log the number for the first one, but the 2nd one is Movie id: 1830630V1 It looks very obvious to me that there is a track in the lower left corner (the other one seemed as obvious as this too) but my sensitivity was marked down greatly after marking both of them. Am I doing something wrong? I have only been dusting for a short time....

[hurons84]

I'm struggling with the same thing. 11 for 11 on specificity but 0 for 7 on sensitivity. The "tracks" appeared obvious, but came up incorrect. I knew it was going to be harder this time around, but I didn't think I was that bad! It might be helpful to have a small block of like 10 CMs that clearly present some of the tracks used for the sensitivity side of this. This was an example that I really don't understand: 968036V1

[jsmaje]

heyjude and hurons84 - I'd most definitely also have clicked on your tracks in 1830630V1 and 968036V1, but did you click on the lowest visible part/particle, as now required according to the tutorial, rather than anywhere along the putative track as was adequate before?

Personally I'm having quite the opposite problem! Having got through 56 movies at my first attempt today, my score amounts to a magnificent 1, having reached minus 3 points within even the first 15 for not noticing tracks that were alledgedly there. Sadly I still couldn't see anything on reviewing the respective movies.
A long-asked-for improvement to the VM experience during phases 1 & 2, when reviewing such missed movies, was an indication of where one should have clicked. For whatever reason we are still required to remain in the dark even in phase 3.

I just hope we aren't witnessing the same initial glitches that occurred at the beginning of both of the earlier phases, where many track coordinates had been wrongly registered by the team, and which took them some time to acknowledge and correct, causing much gnashing of dusters' teeth !

Meanwhile, may I urge H & h above, and all other forum users, to post movie links as hyperlinks that readers can simply click on to view, rather than as plain text. It's easy:
Where required in your post, click the URL button, then paste the relevant address between the url tags, substituting the movie number for the dots:
**stardustathome.ssl.berkeley.edu/myevents_viewer.php?id= ...... [for 'reals']
**stardustathome.ssl.berkeley.edu/ss_virtual_microscope.php?ondeck_movie= ...... [for 'calibrations']
For example, http://stardustathome.ssl.berkeley.edu/ ... e=968036V1.

NB: To prevent the forum or your browser interpreting the above blank addresses as real links (which on clicking could otherwise lead you into an inescapable cyberspace black hole), ** stands for the usual required prefix of http://, which should be used for complete reliability.

Better still though, and much neater, simply re-format as [url = **the address + movie no.] the movie no. again [/url], which will result for example in: 968036V1, as at the beginning of this post.

Note that one must be logged into the site to review calibration movies in this manner, whilst it is unnecessary for the others.

John

[heyjude]

Thanks for that excellent reply John. Perhaps I have been caught out by that problem... I admit I didn't really understand what they meant by clicking on the lowest bit and it still isn't all that clear to me. It was repeated over and over... I think I'll look at the tutorials again and see if I can make sense of it. I'd love some very clear examples as I think I am just clicking 'on' the track.. perhaps in the middle ... and that must be wrong.

And thanks very much for showing how to make the link... sorry for that. I'll make sure I always do in future

[heyjude]

Wow... ok, so I went over the tutorial again and I hope I am not too stupid in understanding it to mean basically you mark the 'end' of the track.. the tail... as far deep into the gel as you can see. So... that is what I am trying to do now. Not too successfully though. I've had two calibration movies mark me down a couple of % points which really makes me wonder. The last one was this: 2941696V1 As you can see, the track is very obvious on the lower left hand side. I marked the tail at the visible end as far as I could see once I was focussed down. Boom... down goes my score. Just like the last one!

So.... either I am misinterpreting what 'deepest part' means... or something is not right. I will continue marking at the tail end and just ignore my score as that's not really the point (I'm here to find dust not win the score contest!), but it does disappoint me that I might be getting discounted as a valid duster because of my score.

[heyjude]

*Another* one.... marked down on this one as well! It's so obvious! I mark the end of the tail upon focusing down... anyone?? What's up?

[straydog]

*Another* one.... marked down on this one as well! It's so obvious! I mark the end of the tail upon focusing down... anyone?? What's up?

I am having the same problem heyjude. I lost points on both of these. The second one in particular is quite obvious.

http://stardustathome.ssl.berkeley.edu/ ... ds=?134,34
214/348

http://stardustathome.ssl.berkeley.edu/ ... rds=?-1,-1
337/24

[DanZ]

Hi Folks,

Thanks for bringing these to our attention. Here's what I've got so far.

straydog - you seem to be clicking the wrong end of the track.

Everyone else - there seem to be tracks that go off in an obvious direction, but in actuality (if you look very closely), the true track goes off in almost (although not exactly) the opposite direction. These are the first of these tracks that I'm seeing. I will investigate and get back to you. In the meantime, try not to click on the most obvious end of the track and see if you can see anything going in the other direction (perhaps a little fainter).

Dan

[jsmaje]

heyjude - there seems to be uncanny resemblance between the tracks in your first example 2941696V1 and the second 8409525V1 (A & B respectively below). I guess these are in fact the same calibration track that has simply been pasted at different orientations. By the way, the quality of the team's pasting has improved beyond recognition since the laughably obvious cals in phase 1!



I should of course have used the term 'deepest' rather than 'lowest' in my previous post (which, given that we are interested in finding more-or-less vertically-aligned 'midnight' tracks, might be mistaken to mean graphically 'lower down' the screen in 2D, rather than the 'deepest' part of the track in 3D).

Meanwhile, in common also with straydog, despite endlessly focussing up and down, due in part to their shallow trajectories and the various confusing optical focussing artefacts, I confess I'm not finding it that easy to decide which 'end' of these tracks is necessarily the deeper. I take Dan's point that one shouldn't go for the most prominent feature, such as the largest blob in fig.B for example, but which of the queried ends is in fact the deeper? My best guess is the sub-particle to the left, but as much on the basis of the expected pattern of a fragmenting impactor.

Am I right Dan? Also, what radius of tolerance are you now allowing for clicking on the 'correct' spot? But why should it matter that much anyway, as long as dusters can demonstrate that they are able to identify tracks of interest somewhere along their length? If someone identifies a genuine track, but happens to have clicked on the 'wrong end', will you ignore it? I hope not.

John

[greuti]

Hi,

Regarding the deepest point of such a track it should be a simple business. It have to be the particle (or at least that one of the remains that entered farthermost) that caused the hole in the collecting tray, i.e. in the aerogel that is inserted in the tray. http://curator.jsc.nasa.gov/stardust/sa ... _Index.cfm

BTW I've had no incorrect counting by marking the large blob in fig.A or B. It is the same high angle track. Currently (it looks like) there are two different tracks in use in the calibration movies.

Peter

[DanZ]

Also, what radius of tolerance are you now allowing for clicking on the 'correct' spot?

I've been trying to find that out from the techs myself, and will let you know.

But why should it matter that much anyway, as long as dusters can demonstrate that they are able to identify tracks of interest somewhere along their length? If someone identifies a genuine track, but happens to have clicked on the 'wrong end', will you ignore it? I hope not.

Yes, I would also assume that if there are enough clicks along a track (regardless of where on the track), the track will get looked at. Nonetheless, this still doesn't help with people's disappointment in getting calibrations wrong. And so, ....

but which of the queried ends is in fact the deeper?

I think Peter's post has answered that question and I'm glad to hear he has not been getting incorrect scores. Let's try to follow his lead and see if that solves some problems.

Dan

[DanZ]

Everyone else - there seem to be tracks that go off in an obvious direction, but in actuality (if you look very closely), the true track goes off in almost (although not exactly) the opposite direction. These are the first of these tracks that I'm seeing. I will investigate and get back to you. In the meantime, try not to click on the most obvious end of the track and see if you can see anything going in the other direction (perhaps a little fainter).

Okay, so in fact, these all seem to be a track that fragmented. And Peter does seem to be getting them right, so I suggest following his lead, including his examples of terminal particles, such as this one. What you may not be seeing (and think you should), is the image we provided at the beginning of the tutorial where the "entry point" is open and at the surface.



With focusing, you may discover you are seeing terminal particles. Think in that way and see how it goes, then report your successes or failures here so we can all see how everyone is doing.

Thanks!

Dan

[DanZ]

*Another* one.... marked down on this one as well! It's so obvious! I mark the end of the tail upon focusing down... anyone?? What's up?

Hey heyjude,

I think if you start trying the other end, you'll have better luck. Give it a go for awhile, and then let me know if things improve for you.

Thanks!

Dan

[heyjude]

Thanks all for your advice. I'll do my best to follow it. It does feel that the specificity of the area required is a little counter-intuitive (wouldn't it just be most important that we identify a track... wherever we get it?) but anyway... good to get all the prompt feedback from the team I am new to this and loving it! I have recently had to give up a career in feature film visual effects due to ill health, so this is a wonderful place to channel my need to pore over pixels!

[hurons84]

When I was using Google Chrome I was 0 for 9. Last night I switched to IE and have been having much more success. My sensitivity is now at 26% in spite of the poor start. You may want to try a different browser to see if that helps.