Changes In This Year's Best Of Photojournalism Competition

Jan 9, 2006

By Stephen Sample & Jared Haworth

DURHAM, NC – The first year of the Best Of Photojournalism contest was remarkable. In the space of a month, we created an international photojournalism contest that — unlike the other major photojournalism contests such as World Press and POYi — covers all of photojournalism, whether presented in print, on television, or on the Web.

Five years ago during that first BOP contest, we took in and judged nearly 25,000 images. Considering that the contest committee members themselves had predicted 10,000 images in the first year, the initial success of the contest was remarkable. And the contest has grown and prospered since: the number of entries grew by a whopping 66 percent from 2002 to 2005. However, the very success of the contest means that it’s now time to revisit some of the entry and judging procedures. We need to make sure that they will still work efficiently with 50,000 images – or 100,000, for that matter.

There are some changes in the entry and judging processes for the 2006 Best Of Photojournalism competition, with the largest changes being in the Photo Editing and Still Photography contests (but all the divisions benefited from the new development efforts).

Registering for contest. All four divisions of the contest are now using the registration system that was developed for the Still Photo division last year. Each entrant fills out the Contest Registration Form and gets their unique Contest Identification Code, which is then submitted along with each entry. This ensures that every entry can be clearly identified, while keeping judging anonymous.

The change to a single online registration system also means that (1) users don’t have to fill out contact information for every entry, and (2) users who enter more than one division of the contest (for example, Photo Editing and Web Sites) can use the same Contest Identification Code for all their entries.

Photo editing division. Starting in the 2006 contest, entries in Newspaper and Magazine Picture Editing categories must be submitted in Acrobat .PDF format: only the Best Use of Photography categories (EU01, EU02 and EU03) will use hard copy. This change should make the entry process faster and easier for everyone, especially for those outside the United States.

The adoption of a digital entry workflow for the Picture Editing division also means that entrants can now enter electronically, using FTP. This offers faster turnaround times, and faster confirmation of submissions.

Still photo division. The Still Photo division has always included a lot of information with each image — story name, headline, caption, photographer and publication credits, copyright, time and location where the image was created, contest category, and more. But that breadth of information also means that there is a lot of room for typos. And some typos (in the contest ID or the category, for example) can get an entry disqualified.

On the entry preparation side, one solution is to make it easier to apply file information like categories, contest IDs, and story names to a group of images. Most of the category errors last year were from images that had been entered both in a single-image category and as part of a story, and which still had the category information from the other time they were entered. Using Photo Mechanic for the captioning process should help with this, since it allows the user to apply headers from an IPTC Stationery Pad to groups of files — and does so without touching the image data, so setting file information in multiple passes won’t result in a lower-quality image.

But even with better user control and validation of the file information, there will be some errors, which will need to be corrected after the images are sent.

Last year, we did some simple validation on the entry data, but the real safeguard was the preview period: users could check their entries online, and contact the Contest Coordinator with any corrections. This worked, but it left all the changes in the hands of one person, with all the time constraints that implies.

Validating headers. While all the information included as part of the Still Photo entry images provides additional opportunities for errors, it also provides ways to check for them. For example, entrants arrange their images into folders by category, and if the category name from the folder doesn’t match the category name from the file headers, that’s a red flag. Similar checks exist if the Contest ID from the Object field doesn’t match the one from the folder structure, or if the eMail address doesn’t match the Contest ID. This year, the import processing code is being significantly smarter about suggesting likely values when it finds a conflict — and those automatic corrections will be noted in the entry confirmation eMail in case the user wants to override.

Still photo & photo editing. This year, both the Still Photo and Photo Editing divisions will allow some of the entry data to be edited directly by the contestant. So if an entry is mis-labeled, it can be corrected before judging begins. On the Editing side, only the credits can be changed; but the Still Photography division allows headlines, captions, credits, and more to be edited.

Editing will be conducted via a secure login interface, using the password the user set up when they joined the contest — so choosing a good password is important. But to guard against bad password choices, or users who leave a shared machine logged in after editing an image caption, changes will be made to a copy of the image database, rather than to the master; and all changes will be reported to the entrant via eMail. So if an unauthorized change were made, it could easily be “rolled back.”

Entrants will be informed when the editing period begins. The editing period will end before the images become available for public view, so once the contest entry archive has been “published,” the information will remain stable.

Some might argue that user editing works against the educational goals of the contest — after all, why bother to get your category right in the first place if you can just go in and correct it later? Whatever happened to taking care with your entries, and responsibility for your errors? And indeed, the contest committee has taken a similar position with respect to captions: by publishing captions “as is,” we are pushing photographers to learn to write — or at least, to learn to write captions. But it was felt that allowing entrants to correct their files prior to the start of judging was a useful educational goal on its own. And the administrative benefits of removing the editing bottleneck are considerable.

Still photos & online judging. Judging a thousand images per hour is grueling work. Some images are clearly not going to be winners, but deciding which image is the best can be a lot harder. As the pool gets smaller, the judges have to argue the relative merits of each entry, and the judging slows down. But having the judges spend more time at Poynter (and take more time off work to do so) really isn’t practical. So as the contest grows, it becomes harder and harder to get through all the judging in the time available. To alleviate this time crunch, the judging for the 2006 contest is being split in half. The judges will make a first “triage” pass over the images remotely, before they go to Poynter, and they then can spend the full week of on-site judging deciding which of the remaining images are really the best.

One issue for the judging in the past has been thumbnail image quality. Using a better scaling method would make it easier to judge images in Gallery view, but even then a larger size thumbnail is needed. The situation for the Single-image view was in some ways even worse, since at standard screen sizes the image was being scaled on the fly, which almost requires a low-quality algorithm.

However, high-quality scaling methods such as Lanczos and Mitchell are slow. Since we are creating multiple thumbnails for each image, performance is critical: there will be over 300,000 thumbnails generated from the Still entries this year, and at two seconds per thumbnail, just generating the thumbnails — without any of the other entry processing — would take a week.

Creating thumbnails. Fortunately, Apple included a Lanczos scaler in Core Image, which is a part of Mac OS X 10.4 (Tiger). Core Image is a graphics manipulation library that automatically optimizes your code to run on whatever processors are available — including the one on your graphics card. Just handing the thumbnail generation off to Core Image — with no other optimizations — cut 40 percent off the thumbnail generation time. And that’s with an Unsharp Mask and a crop thrown in to give the images clean edges. That’s almost three days shaved off the entry processing.

What if? Critical systems, as anyone working with technology knows, are prone to failure. The Still Photography contest judges have tens of thousands of images to judge in a little less than a week: they can’t afford downtime. So, what happens to the Best of Photojournalism competition if the contest server fails during a judging session at the Poynter Institute? Hopefully, very little.

The main Still Photo contest server is kept on-site in St. Petersburg, so that the judges can zip through the images without worrying about network speed. But there is also a second “backup” server with a mirror image of all the contest data that is held in reserve at another location and can be sent to Poynter if needed.

In the past, this was adequate: we could have completed judging even with a 12-hour “hole” while the backup server was in transit (or conducted the judging over the Internet, if we were near the end of the contest). But with the anticipated size of this year’s contest, even overnight shipping takes too long: if we lost eight hours of judging, we wouldn’t be able to process all the images. So we are adding a third backup that can be accessed remotely while the second server is in transit.

Judging results will be synchronized in real time between the primary contest server and a local database in the judging room. If the on-site contest server fails, we will have a real-time record of which images have been voted in or out. That snapshot can then be applied to the two backup machines, and judging will switch over to the online server until the reserve server arrives on site.

For links to the rules and guidelines for each division, please click here. For more information, please contact Kenniff at contests@nppa.org. The Best Of Photojournalism 2006 contest is sponsored by Canon.

Read earlier story.