Slashdot Mirror
1,028,000 Digital Photographs
cdneng2 writes "Rob Galbraith has an in-depth article on the digital
photo process of Sports Illustrated. The article walks through SI's digital workflow of Super Bowl XXXVIII as it sorts through the 16,183 digital pictures shot by eleven of
the magazine's staff photographers and the process all the way to the cover of the magazine. Sorry, no Janet Jackson or swimsuit pics in this article."
[note: this could be construed as a plug. Perhaps it even is, re-reading. You have been warned]
:-)
:-) You could do things like drag an image out of IE/Moz and drop into 'Shake', with Shake being instructed to load the real footage not the proxy version you were looking at in the browser - this image-based-project-load alone saved enormous time when you're dealing with millions of images.
My company (7 of us in total) wrote an asset management system used on a major film in a previous life (we were called 'unique-id' then). We were given the option of being paid and not disclosing the film, or not being paid and letting everyone know which one. It was a *big* film - we took the getting paid option, so you'll have to guess which
The rushes coming in totalled 40 DTF tapes per working day over several months, several hundred million images in all. The same system was used on the 'The world was not enough' trailer, where the large quantities of mostly-naked women
gyrating around with oil being poured on them suddenly made the visualisation tools *far* better than they used to be...
Every image (every frame) was accessible and searchable, notes could be made and a proxy version played back over the net. It was completely automated - logging was done by simply untarring the data-tape or playing the rfid-labelled video tape, with metadata being inferred from path names or rfid tag, all very simple and very effective. Everything was written using OSS tools, mainly PHP and MySQL (and yes, we paid for our MySQL licences
Simon.
Physicists get Hadrons!
I was hoping from the article name that this was going to be about a great Open Content digital photo archive, like PDPhoto, OpenPhoto, or all the great stuff at the Internet Archive or Common Content.
Instead it's about somebody else's photos I can't use. Zzzzzzzzzzzz.
Evan Prodromou | evan@prodromou.name | http://evan.prodromou.name/
Forget the ???
1. Make software that does both
2. Sell to SI
3. Profit!
Under capitalism man exploits man. Under communism it's the other way around.
I'd be interested to see how pro pics taken at the Super Bowl with Nikon's D2H camera (high-frame-rate, 4MP) compare with the digital EOS, especially since the photo editor claims that most of what he's getting from the EOS users is 'shit'.
I'd have a personalized plate on my car, but "toxic bachelor" won't fit into 7 letters.
In all honesty I don't even remember who played in the Super Bowl this year, but I remember Janet Jackson had a nipple shield on. I feel sorry for Boston or Dallas or whoever won the Super Bowl this year.. it'll go down in history as the event where Janet Jackson showed her old floppy boob rather than where some team won their championship.
I'm surprised Sports Illustrated uses relatively cheap hardware and software to edit their photos. ACDSee (Fire!) is $50, and they use some pretty standard (and not all that impressive) IBM laptops for most of the field action. Even at the studios in New York they're using dual 450 G4's. No dual G5's yet. Anyone know how much that camera costs?
What an odd post.
First of all, what "studios in New York"? I work in New York for a creative company, and we mostly use PC's. If someone prefers to work on Mac we issue them a Mac. Same as SI. Most people, in all honesty, prefer to work on PC's at my company, so that's what we give them. And those people work no more slowly than those using Macs (dual 450 CPU's is hardly impressive these days either, so it's a little weird that you'd put that up against the IBM T40's and dual Xeons SI is using. The last Mac we issued was a dual 1.8 G5).
Secondly, what the hell does cost have to do with anything in qualitative terms? It's any company's responsibility to be efficient in budgeting, and part of that means choosing the cheapest tools you can that will reliably get the job done (key word being "reliably" - it's no use spending the least money you can if what you buy is going to be broken half the time). IBM Thinkpads seem a perfectly sensible idea to me for what SI is doing with them - they're reliable, they're not expensive, they're small and easily transportable, and with Pentium-M chips and 768MB of RAM they're more than adequate for what SI is using them for, which is downloading and tranferring image files. This is efficient use of tools.
Similarly, did you even read why they're using ACDSee? We use it at my company as well. It's simply a very fast image viewer; there's nothing I know of that's faster either, or more suitable to the task of sifting through large quantities of images in as quick a time as possible. We use it for the exact same purpose.
I'm honestly impressed at how efficient and organized it seems SI is running their image processing program. They seem to know what they're doing and they've selected the right tools for the job. Who cares if they use "cheap" cameras and PC's? You got a problem with the technical image quality on any of their recent covers?
In workshops for bird or wildlife photography, I think 20 rolls/day is a typical estimate, and a lot of your time is spent finding subjects, or waiting for them to do something interesting, or waiting for the light to be right.
For the Super Bowl, the numbers come out to 40 rolls per photographer. That sounds about right to me. Figure they're getting every bit of every play that they can see from their position, and are shooting 5 frames per second or so.
This sounds like a lot, but for those too lazy for the math....16,183/11 photographers/6 hours of shooting.... it *only* comes out to 4 pictures per minute. When you consider A lot of them are pictures in random succession like 10 in a row, searching for the ultimate still-frame it isn't really as mind boggling as the initial large number seems.
Sorry, no Janet Jackson or swimsuit pics in this article.
Now as far as that. How many other geeks out there are for Sports Illustrated starting a SETI-like distributed network program for their photos? Imagine the SI photographers taking thousands of swimsuit shots and sending them off to your computer for you to "process". Count me in.
I did read the article.
It's clear that if ACDSee can't show 11MP JPEGs fast enough for the editor on modern hardware then it's the wrong tool for the job. Sports Illustrated wrote a custom app to transfer images from the cameras (I've written one myself and it looks and functions in a very similar manner) then they need to write one that will pre-cache more than one image ahead. It could also do tricks like show the next three images as small thumbnails on the bottom of the screen, letting the editor quickly skip the shots of the photographers foot. No matter, a 3GHz machine is way more than enough hardware to show large JPEGs at 3 frames a second.
I work (in IT) with photographers/journalists at a small news organization. True, they don't need the bells-n-whistles of Photoshop (effect filters, etc.). And yes, cropping is probably used on every picture that goes in the newspaper.
But:
Tempo. Photographers storming in from say a fire, an accident, etc., with a large number of images, and with minutes until the next assignment, or until the pages goes to print. Every second often counts, many times a week.
Good hardware shaving off a little time here and there helps.
Image corrections. Pretty much every image (to be published - in print or online) needs sharpening, color/hue/brightness corrections, and similar touch-ups (but never image manipulations like cloning, etc. Then, like you say, it would cease to be a journalistic photo, and become misleading and/or fraudulent). Actually, even cropping is done with caution - does the meaning change? is it still representative of the events that took place? etc.
Photoshop is still among the best software for doing this (corrections). Plus, most photographers are familiar with it, so freelancers, temps, etc., can jump into the production chain and be productive immediately.
So, I guess I'm arguing that they do need fast computers (we use P4s on the desktop, SUN servers & SAN) and Photoshop (we have additional image software, for batch conversions, etc.).
668.5
Anyone know how much that camera costs?
The cameras are (1D)s (parenthesis because the 1Ds is a different camera costing twice as much). They've come down a bit in price now that there's a couple successors to it out, but I think you're still looking at between $3500 and $4500 for the body. The camera's pretty crazy -- up to eight frames a second means it rocks for sports and stuff.
Don't forget, though, that this is the camera body. And in sports, you need a really fast lens if you don't want a big blur -- figure several thousand more for the lens. (And a lot of photographers -- I'm not sure about SI -- carry around multiple lenses. Actually, the really good ones have a camera for each lens, and keep them on a strap, so you drop one camera and pick up one with a longer lens when you need it. Costs rack up real quick.)
By way of being a total 'photogeek,' there are a pair of cameras that 'replace' the 1D now: the 1D Mark II (8.5 fps, 6 megapixels?) for around $4500, and the 1Ds at 11 megapixels (though not nearly as many frames/second as the others), going for about $8,000.
I'm not sure what your point is, though. Why spend $5,000 on a laptop if a $2,000 one works just as well?
________________________________________________
suwain_2
Unfortunately, take a look at that file name. The Canon files repeat every 10,000 images. If we don't rename the files, we eventually end up with multiple images for the same filename. This is unfortunate. However, it is a heck of a lot better than the Nikon scheme.
A single image taken on any Nikon pro camera has the file name of DSC_0001.JPG and DSC_0001.NEF. This means that when you put the files in the same folder, they end up overwriting each other. This is unfortunate. On the newest camera (D2H), they added support for setting an EXIF field to identify the photographer. This means I can figure out who shot those photos as well. You cannot do this with the D1X or D1H.
So in all cases, we really want to rename the files.
We do keep the date information. Unfortunately, it is not always set correctly. But even if the date were set correctly, we still get the files out of order; we just get bags of digital cards as the game progresses. We would have to delay the edit of the files if we wanted to rename the files based on time.
But let's pretend ACDSee did sort the images by time. This is actually not really what the editor wants. Keep in mind we have 11 photographers shooting at once. The editor wants to be able to see the same sequence from the same photographer before seeing the other photos. In other words, if we have photographers A, B, C, and D, we want the order of photos for a specific play to be AAAAABBBBBCCCCCDDDDD, not ABCDABCDABCDABCD.
--Sam
Not true. You can get Canon L glass on some of the middle-range lenses for half that. The problem is that middle-range will not get the kinds of shots SI needs. They need big-honkin' glass to get shots from any distance at a good speed. Just look at all the pretty white lens bodies on the sidelines. The 2.8L 400mm IS USM goes for about $5500.
The funny thing is, they use the same glass for their swimsuit issue to flatten the depth, (though the 300mm is usually enough). Walkie talkies are standard issue when the photog has to instruct the guy holding the reflector on his model to "move a little to the right".
Also, for Canon, RAW != TIFF so maybe the files aren't as big as you're thinking. My 6 megapixel Canon 10-D generates 5-6MB RAW images (they vary in size because they have a jpeg embedded in them).
.RAW files (which ARE compressed) are smaller than uncompressed TIFF files.
They vary in size because they're compressed, mostly.
Further, for an interesting scene (i.e. not with the lens cap on), this is relatively uncompressible data, so even if you could in some way encode the raw CCD data in PNG format, you wouldn't see much of a gain.
That's not true. Even in an interesting scene, you will have large fields of similar colors which can still be compressed losslessly with good results. ZIP won't get you the best results because it doesn't understand how the data is structured. It's just like how FLAC gets much better compression than ZIP for audio, even though both are lossless. If you understand the structure of the data and where the common patterns emerge, you can tune your compression algorithm to those andcan get decent results.
That's why
I don't think that's so true (anymore). Colour management on Mac/PC is practically identical when it comes to Photoshop/Illustrator/InDesign, and colour profiles have been handled better in Windows since 98 (I think, certainly 2000/XP). Yes, Colorsync is still nicer, but there's not as much of "wow, that's so much better than PCs" as there used to be. The historical reason is very much valid though, people work best when they use something they're comfortable with and used to.
c cessories/ sony-artisan.shtml
:)
As for Mac monitors being *completely* accurate out of the box, that's wrong. They're certified to a certain factory level - not wildly inaccurate by any means and *much* better than most standard off-the-shelf components, but still not good enough for critical work (in fact even Apple themselves recommend using hardware calibration on their website). Besides, devices go "out" with time anyway - esp. CRTs, which on the whole are much better at colour than LCDs: though the Apple Cinema displays are lovely I'd much rather have something like a Sony Artisan:
http://luminous-landscape.com/reviews/a
There's no way around it, you MUST profile your devices with a colorimeter (the Monaco equipment is excellent, GretagMacbeth also, and the Pantone Colorvision stuff serviceable and cheap) unless it's got hardware built in such as the Barco's or high-end Sony's, it's the only way to be sure. Also, a good thing about profiling is you can set devices to specifically match each other, so say you've just bought a batch of laptops to go along with the workstations, you can easily profile them to all look exactly the same - very useful when the photographers may have to share resources. I'm not going to start on profiling print devices or the intricacies of open or closed loop colour systems though, cause it all gets very boring