difference between callibration and real movies

[Rude-E]

I've noticed quite a big quality difference between the calibartion movies and the real movies, the calibration movies are of a lesser quality, there are clear steps visible in the colour (grey) gardients.

Has anyone else noticed this?
And could this possibly influence the results? (I think it could)

[icebike]

I've noticed quite a big quality difference between the calibartion movies and the real movies, the calibration movies are of a lesser quality, there are clear gradients visible in the colours.

Has anyone else noticed this?
And could this possibly influence the results? (I think it could)

I'm not seeing this at all. I see the opposite. Calibration movies are almost always properly focused, and have all their frames. That can't be said of all of the real ones.

[DustBuster]

Yes. I'm sure some of the images are 'faked' and you can tell they are calibration movies (among other ways). You just need to check it for a particle track and mark it accordingly. The fact that it is a calibration movie shouldn't have any effect on one's ability to determine if it has a track or not.

[cldrjones2]

The particles on a calibration movie are bigger than the real ones, at least that's my experiance.

[Rude-E]

I'm not seeing this at all. I see the opposite. Calibration movies are almost always properly focused, and have all their frames. That can't be said of all of the real ones.

What I mean to say is that when the background (the actual gel) goes from a darker grey on one side of the image to a lighter grey on the other side of the image, this happens in visible steps, but the real movies do not show this, the change is smooth.

[fjgiie]

Hi Bryan Mendez,

Six times on the same calibration movie with the large track easily visible, when you click the hole in the doughnut, it registers as a miss.

A miss means that this is not a calibration movie, but it is indistinguishable from the other large tracks.

So, there is a calibration movie out there, that if you click the track, it scores a miss.

Thanks,
fjgiie
PS. I did not miss the hole in the track three times.

Edit: I came across it again and missed again...10689B, and 10635A
Thanks

[slindemann]

I think Rude-E mean that the calibration movies are more compressed and therefore bigger area with approx. the same color will get the same color in the view.
I have notice the same when I am getting a calibration movie.

[bmendez]

This is an effect known as "contouring." It is a result of lossy compression algorithims such as is used in JPEG compression (which we are using in this project). It is indeed visible in some of the calibration movies, and I have seen it in some real movies too.

-Bryan

[carmencr]

This is an effect known as "contouring." It is a result of lossy compression algorithims such as is used in JPEG compression (which we are using in this project). It is indeed visible in some of the calibration movies, and I have seen it in some real movies too.

-Bryan

Now that I have a few thousand movies under my belt, I am able to identify calibration movies that will have tracks about 80-90% of the time, and calibration movies without tracks perhaps 50% of the time. (Usually, the latter are ones I think will have tracks, but don't. That's still a correct ident of the movie though.) This is based purely on first image load, before I mouse over the focus control at all.

This is entirely due to the countouring artifacts. It's quite easy to see at this point that most of the real movies are much sharper than the average calibration movie. It's not a hard-and-fast rule, but it serves me well 75% of the time...which is kind of defeating the purpose of the calibration movies, isn't it? I'm not so much identifying tracks at this point as identifying features of the field, and knowing that I have to look more closely for a track. It's subconscious, and I try not to get into that patern, but when you're doing hundreds a day it will start to happen without thought.

Despite the fact that it would affect scores, I would suggest that the next batch of calibration movies (assuming more get made to keep people on their toes) be carefully selected and compressed to avoid this.

[ToSeek]

I agree - I definitely feel as if I'm being trained to recognize JPEG compression artifacts more than tracks. About 80% of the time I can tell from the get-go if it's a calibration movie. (In addition, I've found that if a movie loads extremely quickly, i.e., it's still in my browser cache from the last time, it's almost always a calibration movie.)

[erazorius]

I found one interesting thing about the movie photos. Almost all movies with track have strange layers on photos. I used paintshop Solarize function to highlight the borders of the layers. "No track" movies don't have these layers (photos are smooth).
http://img220.imageshack.us/my.php?image=track1fn8.jpg
Marked with dots
http://img88.imageshack.us/my.php?image=5wi7.jpg
Also i find that photos with trace are size of 2-10KB, and photos without trace
are of size 20-25KB.