Meta-Metadata

By Marc Rochkind

At first there were only the images themselves, maybe with some shooting or processing data about them if the photographer remembered to record it. Ansel Adams was careful to record exposure, but had an "unfortunate disregard for the dates of [his] negatives" and an "anti-date complex" (his words, in Examples: The Making of 40 Photographs, pp. 42–43; many years after Moonrise was taken an astronomer calculated its time as 4:05 p.m. on Oct. 31st, 1941).

With a digital camera, it's as hard as ever to take pictures as good as Adams', but it's a lot easier to record the shooting data. The camera stores it automatically inside the image file. This so-called EXIF data is data about image data (the pixels themselves), or metadata. Most viewing applications display metadata for an image, and so do websites like SmugMug and Flickr.

Cataloging applications like Lightroom, iView/Expression Media, and Aperture go much further: They maintain a database of images, including image metadata. That's data about data about data, or meta-metadata. You can see a bit of this in Lightroom if you open the Metadata Browser (Library module, bottom left), which lists interesting facts like the number of images for each camera and lens.

Even better, Lightroom (and Aperture) use the free and open SQLite3 database, so it's very easy (for a programmer, anyway) to access their catalogs. Although Adobe doesn't document the structure of Lightroom's database, it's not hard to figure out. You can use free utilities like SQLite Database Browser (sourceforge.net/projects/sqlitebrowser) to look at the structure of the database tables and the data that's in them.

That's what I did to develop a small application I call ImageReporter. It shows a lot more than Lightroom's Metadata Browser (such as the average aspect ratio after cropping), and it also lets you filter the report by type of image (DNG, JPG, etc.), by rating (1 star or more, etc.), and by time interval (last 30 days, last 90 days, etc.). You can get a free copy of ImageReporter for the Mac or Windows at ImageIngester.com. If you just want to look at a sample report, I've posted one of my recent ones here.

It's fun to look at the metadata summaries in Lightroom's Metadata Browser or with ImageReporter, but it's even more interesting to discover things about your photography from them. I'll give two examples of what I just learned.

First, from ImageReporter's section on focal length, which shows what focal lengths zoom lenses were set to:

662	NIKON CORPORATION / NIKON D200 / 17.0-55.0 mm f/2.8
   245    37%	 20mm
   152    22%	 30mm
   109    16%	 40mm
    43     6%	 50mm
   113    17%	 60mm

This is almost the only lens I use on my D200, and the only zoom. The data (meta-metadata, but I'll keep things short) shows that I'm using the lens as a wide angle. I didn't know that! I probably should think about getting a 12–24 zoom, since I appear to be at mostly wide-angle focal lengths. Maybe I really want to go wider and can't?

Cropping is OK if the alternative is getting wet or showing a beer can. Otherwise, it's better to move your feet.

Second, I'm cropping too much! Look at this data from ImageReporter's "Average Crop of Images That Were Cropped" report:

81%	[unknown]
90%	Canon / 8400F
43%	Canon / Canon EOS DIGITAL REBEL XTi
64%	Canon / Canon PowerShot G9
76%	CASIO COMPUTER CO.,LTD. / EX-S500
55%	FUJIFILM / FinePix S5700 S700
70%	Leica Camera AG / M8 Digital Camera
95%	Minolta / DS Dual4
69%	Minolta / Scan Dual IV
59%	NIKON / E3100
60%	NIKON CORPORATION / NIKON D200
58%	NIKON CORPORATION / NIKON D70
51%	Panasonic / DMC-LX2

With my Leica and its 35mm lens (the only one I use), I'm cropping to 70% (by area). With my D200, I'm cropping even more, to 60%. Not only is that with a zoom, which ought to allow me to fill the frame, but the focal-length data I showed above says that I'm at the wide end of the range. Sure seems like I ought to do more careful visualization before I click the shutter and probably zoom in a bit.

Maybe the problem is the aspect ratio? With a 1.5:1 image size, which is what the Leica M8 and Nikon D200 both use, you're forced to crop if you want a more squarish image. ImageReporter has an aspect ratio report, too, part of which shows:

1.41	Leica Camera AG / M8 Digital Camera
1.46	NIKON CORPORATION / NIKON D200

No, it's not an aspect-ratio issue, because I'm typically cropping to close to 1.5:1 on the D200.

What about my better images? Maybe I'm doing extreme crops on snapshots, but I set up more carefully when I'm really trying? Turning on the 4-star-or-more filter, I see (in part):

78%	Leica Camera AG / M8 Digital Camera
68%	NIKON CORPORATION / NIKON D200

That's a little better: 78% and 68% vs. 70% and 60%. What's more, my better images are less than 1.5:1, especially with my D200, which means there was a method to my cropping:

1.43	Leica Camera AG / M8 Digital Camera
1.27	NIKON CORPORATION / NIKON D200

Still, the data shows that I need to move my feet more when I'm shooting with the Leica, and, with the D200, I need to do that and also zoom in. Maybe I don't need that 12–24mm zoom after all.

You can also use meta-metadata to learn about other photographers. I just got 580 original JPEGs shot by the photographer my niece hired for her wedding last June. I imported them into Lightroom, assigned them to the Quick Collection, and then used an option in ImageReporter that limits its scope just to that collection. Looking at the report, I now know a bit about how that photographer worked.

He used two camera models, a Canon 30D and a 1D Mark II, using the latter for 79% of the shots. Lightroom didn't record the serial numbers, so I don't know if he had one or two or each. Lightroom didn't pick up any lens data, either, so I don't know what lenses he was using. But here's the focal-length data that ImageReporter collected (it rounds to the nearest 10mm, to help the grouping):

123	Canon / Canon EOS 30D
	50    40%	 30mm
	19    15%	 40mm
	 8     6%	 50mm
	11     8%	 60mm
	15    12%	 70mm
	20    16%	200mm
457	Canon / Canon EOS-1D Mark II
       101    22%	 20mm
	38     8%	 30mm
	45     9%	 40mm
	20     4%	 50mm
	18     3%	 60mm
	39     8%	 70mm
	 9     1%	 80mm
	37     8%	 90mm
	20     4%	100mm
	11     2%	110mm
	 8     1%	120mm
	10     2%	130mm
	16     3%	140mm
	14     3%	150mm
	 9     1%	160mm
	11     2%	170mm
	 4     0%	180mm
	 6     1%	190mm
	41     8%	200mm

The two bulges (8%) at 70mm and 90mm make me think that this photographer used two zooms on his 1D (or two 1D cameras), a 24–85mm (rounded to 20mm and 90mm) and a 70mm–200mm. That would make sense for a wedding photographer. If Canon made a 24–200mm, he'd probably buy one.

The 30D (or maybe two of them) looks like it may have had a zoom from around 30mm to 70mm, stopping there because there are no shots between 70mm and 200mm. It's probably not a 24–70mm, because ImageReporter says there were 50 shots taken within 10mm of 30mm and none wider. It must have been Canon's 28–70mm ƒ/2.8. Indeed I found at least one shot made with that camera at 28mm and ƒ/2.8, so it could not have been a slower zoom. The gap from 70mm to 200mm means either that he put a 200mm prime lens on a 30D or he had the 70–200mm on the 30D zoomed all the way in while he had the wider zoom on another body. This makes sense, and is probably part of his normal work pattern during certain parts of the wedding. If I spent more time on the photos themselves I could probably figure this out for sure.

I didn't pay that much attention to the photographer during the wedding, but it's definitely interesting to find out a bit about how he worked from his meta-metadata.

I found out some things about my own shooting, too. I don't need a full-frame 35mm digital, or even a medium-format digital back, at least not yet. All I need to do is try harder to compose in the viewfinder instead of in Lightroom.

___________________

Marc

Comments

[jg]

Brilliant stuff. Especially the things you do vs. things you think you do. Self analysis has never been easier... I'll certainly give it a try. Wish there was a tool like that to evaluate the rest of me!

[Jon Fitch]

I used a EXIF tool that works on Nikon NEF's in a folder. It was fascinating to see what my habits were. Almost all of my shots are made at the wide or long end of a zoom. I am basically using zooms as "double primes."

I've been trying to decide if I should just buy high quality primes at the focal lengths I want.

[Bill Mitchell]

"Cropping is OK if the alternative is getting wet or showing a beer can. Otherwise, it's better to move your feet."
That's nonsense!
The content and perspective of an image is determined by the physical relation of the photographer to the subject. Once the desired composition has been achieved, the proper lens should be chosen (or zoomed) to crop out the extraneous material.
Where you stand is an aesthetic concern, cropping is a technical one.
"If you're not close enough, you're pictures are not good enough," only applies to news photography.

[Jason]

This is fascinating stuff, but I do it all day at work. When I get home, analyzing reports is the last thing I want to do. Hopefully I've got enough pics to do one analysis, act on it then never look at that again! :-)

On the other hand, sometimes it's fun to go get that lens you didn't need and found out you really did.

[My Camera World]

Thanks for taking the time and effort to share with us what looks like an interesting tool to help us better understand our shooting methods.

Niels Henriksen

[Matthew Robertson]

Fantastic program. Examining the differences between my 1-star and 2-star images has shown me some statistical trends that I would never have otherwise seen. Thanks so much for putting it together and making it available.

[01af]

Programs that analyse metadata (i. e. meta-metadata generators ;-) are not a new thing. For an example, see here:

http://www.cpr.demon.nl/prog_plotf.html

This little freeware program is around for at least three years now ... (Windows only).

-- Olaf

[Tim]

Taking Bill Miltchell's comment further: maybe dSLR users are having a phase of not caring about perspective? That is a critical area in which you'd expect to see a difference between zoom/crop/moveyerass methods. Tells me people are shooting "a photo of something" with little care to the angles it makes *in the resultant composition*.

I was skeptical when I started reading this; `meta-meta-data would be something like applying Ratings to EXIF, such as "I don't like the fact it recorded this as 28mm"', I thought. And `that's not meta-meta-data, it's a (crude) statistical analysis'. But by the time I got to the bottom, I see Marc's learnt something about his and someone else's styles, with a definite potential for improvement, so I'm warming up to the analysis a bit.

I shoot a D200, some MF and LF work for fun. One thing I notice is when I'm used to images 7000px square at 16-bit colour, I treat every pixel coming off the D200 as precious. So maybe Marc would benefit from dumping the Leica, get a Hassie, shoot square format (see how it affects your crop style having to work with a different source aspect-ratio), and benefit from the slow-down that paying per frame brings, too? ;)

[robert e]

Thank you for that illuminating guided tour of your metadata.

Re the D200/17-55mm data (obviously not being privy to any other knowledge about your shooting): If you're spending 40% of the time at normal and 25% at tele, I'd say you're milking the whole range of that lens, while showing some preference toward wide and normal.

It would be instructive to know crop percentage per focal length, as, looking at just the numbers, you may be wanting more reach at the long end as much as you may be cropping away at the wide shots.

In general, though, is it not better to have some slop to crop away? Especially so with respect to your implication that the better images are getting cropped for common print aspect ratios (which you wouldn't bother to do with those not worthy of printing).

Assuming that that is how you work, and accounting for the D200's slight viewfinder crop plus the cropping necessary for larger print sizes, I submit that averaging 70% per keeper shows rather impressive in-camera framing.

[Thiago Silva]

Geez, I'm a stats junkie. You totally got me with this.

[Ken Tanaka]

Marc,
This is a good topic to highlight. I've long argued with my emulsion-hugging acquaintances that digital photography's EXIF data represents the greatest instructional resource that photography has ever seen. Most still hug emulsion, make narcotic images, and teach photography like they were stuck in the early 20th century. That's fine.

No, studying EXIF data won't turn a crappy photographer into an icon. But at least, if used thoughtfully as your piece suggests, it can help to prevent a crappy photographer from making the same technical mistakes over and over. He can become a technically better crappy photographer in no time at all!

[Luc Novovitch]

May I suggest a short version for this posting?

"I am unable to visualize and frame a decent image in the field: therefore I recrop."

[Claire Senft]

Since when is getting your feet wet a concern?

[Robin P]

"I am basically using zooms as "double primes."
I wonder how many of us this applies to? It's something I began to realise a couple of years ago so now I have to carry more lenses but get better image quality. Last year I took a 70-150mm zoom on holiday for the convenience but am frustrated by knowing Id have got better shots with my 85mm and 135mm lenses.

Cheers, Robin

[peter in bangkok]

my guess: an analysis of 100 photographers "pattern" according to Marc would show most would do with primes, or even only one prime..

[Marc Rochkind]

Some responses from the author:

Criticisms of my photography are surely true--mastery of perspective, as well as everything else. I'm a much better programmer than I will ever be a photographer!

The article was supposed to be about the idea of getting insights from metadata and about the tool. My personal role was only that of a bad example.

Regarding square format: I have noticed that many of my favorite framed pictures are close to square. I do have a Mamiya RB67 (almost square), but haven't yet brought myself to shoot film (again).

As far as getting my feet wet goes, I trust from the picture you can see that it was COLD. ;-)

--Marc

[Ken Tanaka]

""I am basically using zooms as "double primes."
I wonder how many of us this applies to?"

Robin: Probably many of us. I know that when I use a zoom lens I tend to favor certain focal length positions in its range.

But that observation does not necessarily suggest replacing a zoom with a prime. The fact is that I don't generally know what focal length I'll need next. So the the zooms will always win for convenience and simplicity when it comes to impromptu photography. Frankly, for all practical observable purposes, the best quality zoom lenses today are easily the optical rivals of most primes within the mount.

So zoom on!

[Strict]

Thank you for the wonderful work.
I suggest you for a future version that it would be very intresting to get also the data about f-stops analized.

[Claire Senft]

So what if the water was cold. I can see avoiding frost bite though. The number one most important item affecting your photographs is "where was the camera located."

[Robin P]

"Frankly, for all practical observable purposes, the best quality zoom lenses today are easily the optical rivals of most primes within the mount."
Yes Ken, I'd reluctantly go along with that if I was using a DSLR or 1990's film camera but the Konica AR mount cameras I use have wonderful primes and very compromised zooms.
It's back to the old argument that primes tend to be brighter and for anything shorter than about 85mm one can "zoom with the feet".

Cheers, Robin

[John Banister]

Thanks very much for writing (and writing about) ImageReporter and making it available. I'd pay $20 for a program that would let me edit EXIF data, so that I could enter the same sort of data a digital camera records for images scanned from film and then be able to similarly sort those. I know I can record that information as an EXIF comment or an IPTC description, but that's not compatible with the sorts of things a program like ImageReporter does.

[Marc Rochkind]

John--

Actually, ImageReporter takes its data from Lightroom's catalog (database), not directly from EXIF data in image files. So, if you enter data into Lightroom's metadata panel, ImageReporter will pick it up. Lightroom will use it, too, for sorting, finding, etc.

--Marc

[ania]

Now that's quite intriguing :)

[alan]

ImageReporter is a great tool. Thanks, Marc! It's much more user-friendly than the combination of sqlite3, awk, and R scripts that I've been playing with to find and summarize/plot much the same information.

I wonder if you've made any changes for Lightroom 2? The "Capture Time" setting doesn't appear to work for my LR 2 catalogs -- no images are found when I specify a value there.