Data Findings Interactive

Objective Photo Ratings Compared Across EXIF Data Categories

March 16, 2010

Hello fellow photographers and photo enthusiasts! I've completed some preliminary data mining on the relationship between a photo's objective rating and various categories of EXIF data. With the backing of those data results, I can finally answer age old (and somewhat non-sensical) questions like "Are Nikons or Canons taking better pictures?", "At what focal lengths are photos the best looking?", and "Is your manual white balance setting really making you better than those auto setting users?"

Please keep in mind that correlation does not imply causation as well as the sample methodology and size used by this report (discussed at the end of the page). Now, let's dive right in.

Nikon users take better photos than Canon users (especially the corporate version?)

In the above graph, you can see that Nikon was the top make in terms of average photo rating handily beating Canon which took the second place (making me sad as a longtime Canon owner), while Apple was the worst. This makes sense as Apple's camera offering involves mostly phone cameras and people tend to take more snapshots than compositions. But is Nikon the definitive champion here? The conclusion may seem slightly muddled by the fact that Nikon for some reason shows up in 2 categories in EXIF data, both as Nikon Corporation and as just Nikon. The majority of the photos in the sample were in the Nikon Corporation category while only a few were in the Nikon category (you can click on the graph source link and drill down on the table for the data behind it). If the two categories for Nikon were combined, the combined Nikon average would still beat Canon.

One possible explanation for Nikon beating Canon can be that Canon's point and shoots might be more popular than Nikon's and thus pulls down Canon's average rating by having a higher proportion of snapshots. A Canon SLR owner like myself can cling to the hope that if we separated the data out the SLRs might well stand their ground.

Unfortunately, Nikon D90 users dispel any such notion in the following graph:

Highest Average Ratings by Camera Models: D90 reigns supreme amongst SLRs

There is definitely a positive correlation between how much you spend on your camera and how well your photos rate. Photos taken with the D90 however, averaged better ratings than the pricier D300.

Lowest Average Ratings by Camera Models: All Point and Shoots

No surprises here. The lowest ratings belong to those using point and shoots.

The Effects of Camera/Lens Settings

Another potentially interesting area was to check to see if there were any patterns when examining things like aperture, shutter speed, iso, white balance, and focal length.

Aperture

Here, each faint red dot represents one photograph and the solid red dots represent the average rating of all photos for the given aperture. The X-axis shows the F stop numbers, so the left hand side represent wide open apertures and the right hand side represent smaller apertures.

First of all, it looks like most photos in the sample were taken with an aperture between F/2.8 to F/5.6, and that segment also represents the lowest rating averages. Simply using an aperture outside the F2.8 to F5.6 range almost guarantees that the photo will not be terrible ( < 1200 ). Outside that range though, there does not appear to be a strong correlation between aperture and photo rating.

Shutter Speed

Going into this exercise, I would have expected some kind of an anomaly in the graph around the 1/60 sec mark due to the fact that around 1/60 sec is where camera motion blur starts to become a problem. Instead the graph shows a rather gradual decline of ratings as the shutter speed gets slower.

ISO Speed

The X axis shows ISO speed. There really does not appear to be a correlation between ISO setting used and the photo ratings.

White Balance Mode

Looks like on average, it is simply better to leave your white balance setting on auto. Unfortunately there was not enough data for other categories such as Tungsten and Fluorescent to show up.

Focal Length

This graph is currently only showing the obvious, which is that point and shoots and camera phones which have a very small focal length average lower ratings than other focal lengths which most likely represent DSLRs.

Post Processing

Post Processing Software with the Best Ratings

Using a recent version of Photoshop seems to be a very strong predictor for high ratings and yes, Mac is beating out Windows.

Post Processing Software with the Worst Ratings

I did not know you could use QuickTime to edit photos. The lower ratings tend to be older versions of software here or software that comes with your camera.

Creative Commons photos average much higher than All Rights Reserved

Flickr very successfully pushed the Creative Commons licenses for photos. Although this is not an EXIF related data category, I thought it would be interesting to see the ratings breakdown between photos marked as Creative Commons and photos that are All Rights Reserved.

What do you know! Maybe people who take the effort to make it easier to share their work do tend to take better photos.

Sample Size, Methodology, and We Need More Data!

  • Sample Pool: the total photo sample pool consisted of 866 photos and 695 of them had EXIF info. 800 of the photos were randomly selected from Flickr's public rss feed, and 66 of them were submitted by users. For each individual graph, you can click through to the source table and see the photo counts and underlying data.
  • Comparisons: 42460 comparisons were done by 176 users, but the distribution is rather uneven, so ratings were heavily influenced by users who did more comparisons.
  • Objective Rating: mechanics of EloGrade Ratings

In a nutshell, more data is needed for statistical significance and more interesting analysis, I will revisit these same graphs when more data is available as well as think about what other cool categories can be checked out.

Unless otherwise stated, the content of this page is licensed under Creative Commons Attribution-ShareAlike 3.0 License