[X] My Assistant

[Rakhir]

I don't found any post about Stardust@Home so I created a new thread.

Stardust@Home is a distributed search by volunteers for interstellar dust in the Stardust interstellar dust collector.

Volunteers have to pre-register for an expected start in spring of 2006. You will also have to pass a test to be qualified for the search.
Indeed, unlike the distributed computer projects running in background tasks, this project is using your eyes to scan "focus movies" thanks to a Virtual Microscope.

According to estimations, there should be about 45 interstellar dust impacts in the collector.

Besides the satisfaction of contributing actively to this sample return project, your name will appear as a co-author on the paper announcing the discovery of the particle.

http://stardustathome.ssl.berkeley.edu/

Rakhir

[deglr6328]

I'm having a terrible time with bad focus images. I've done maybe 100 "movies" and at least a third have the surface in focus WAY far away from the center of the slider bar. Some movies are so devoid of any features at all that I can't tell if I'm actually looking at different images or if the javascript has frozen up and I'm stuck on the same one. If we're looking for ~10 tracks in ~1 MILLION images I have little hope the method will deliver real results with this kind of performance. The images should contain a reticle fiducial like this which rests on the srface so the viewer can see immediately where the surface is.

It looks to me like all they've done is get a block of aerogel under the microscope, focus to the surface and tell the autoscanner/imager "ok here's the surface, scan the whole block assuming this level at xx microns will be where the surface is" regarless of surface level deviation, when what they should be doing is surface level calibrating every movie made by using a little piece of software to see where the highest contrast is (like a digital camera does) and or using a small/thin surface fiducial at the image edge. Also Dilo is right about those dots that appear on every image, it makes me think they didn't even flatfield calibrate the CCD on the imager before they started!! not good. On the other hand the user interface is pretty nice and natural to use.

[RobR]

I'm having a terrible time with bad focus images. I've done maybye 100 "movies" and at least a third have the surface in focus WAY far away from the center of the slider bar. Some movies are so devoid of any features at all that I can't tell if I'm actually looking at different images or if the javascript has frozen up and I'm stuck on the same one. If we're looking for ~10 tracks in ~1 MILLION images I have little hope the method will deliver real results with this kind of performance. The images should contain a reticle fiducial like this which rests on the srface so the viewer can see immediately where the surface is.

It looks to me like all they've done is get a block of aerogel under the microscope, focus to the surface and tell the autoscanner/imager "ok here's the surface, scan the whole block assuming this level at xx microns will be where the surface is" regarless of surface level deviation, when what they should be doing is surface level calibrating every movie made by using a little piece of software to see where the highest contrast is (like a digital camera does) and or using a small/thin surface fiducial at the image edge. On the other hand the user interface is pretty nice and natural to use.

Bah, Humbug you say.

I've been 550 'real' slides now and I don't see that big a problem with the poorly focused examples. Just click bad focus and go on. the odds are low but... you might have one named after you!

my stats:

Your Overall Score: 186
Total Movies Viewed: 761
Your Rank: 885 out of 7330
Specificity: 100%
Sensitivity: 89%

Movie 20962 31 5 Indeterminate
Movie 41699 41 9 Indeterminate
Movie 43214 35 11 Indeterminate
Movie 22712 37 3 Indeterminate
Movie 41954 37 7 Indeterminate
Movie 41322 47 7 Indeterminate
Movie 43323 49 5 Indeterminate
Movie 41590 46 10 Indeterminate
Movie 41491 48 1 Indeterminate
Movie 44039 28 4 Indeterminate