Petapixel carried a report on a French startup, Regaind, who have created a computer based photo critique programme called Keegan. The programme, based on input from human photographers, offers a basic critique of your images, together with a mark out of ten and a more detailed breakdown of elements of the picture.
The first thing to be said is that it’s great fun to play with. It has a simple drag and drop interface and only takes a few seconds before delivering its verdict. Beyond that, while Keegan is certainly artificial it is clearly lacking in intelligence, at least, in the sense that that word is normally understood. Though, having said that it is definitely cleverly programmed. It can identify blur, angles of vision, colour intensity, placement of the subject and much else besides and generate criticism of images based on those criteria. However, once you get beyond these basic elements of an image the programme runs into trouble. It is simply incapable of dealing with complexity or subtlety.
Of the images that I tested it with, here is the one that scored highest:
Keegan awarded me a very respectable 9.6 for this image with the comment: ‘Timing is everything, such a great composition, and really pretty perspective. The light is hitting your subject just right…yes and thank you for this interesting image. Very nice job, perfect. Congrats!! You get a 9.6/10.’ Now, this is without a doubt one of the most boring pictures I’ve ever taken and the only reason I did was to test out the abilities of my XF55–200 lens.
By way of contrast, the following shot garnered my lowest scores of 3.6:
Keegan was not impressed (remembering that Keegan is a computer and is therefore incapable of being impressed or of having any other emotional or rational response.) ‘This is what we call a bad composition and I don’t think this is as amazing as you thought… It’s flawed but you’ll get there: keep working on your composition, framing, and lighting. This at most 3.6/10.’
But this is, in fact, quite a good image (in my opinion of course — I’ll leave you to reach your own judgement). Keegan’s problem with it is that it is quite a complex image. There’s a lot going on here and there are multiple subjects and points of interest. The lighting is also complex here with multiple sources and lots of reflections. Because Keegan is not intelligent it is incapable of understanding and interpreting this image.
To be fair to the developers they do point out that Keegan is at a very early stage of development and they also stress that it is aimed not so much at experienced photographers but more at the wider photographic community, the ‘everyday’ photographers. Viewed this way, programmes like this could be potentially useful in giving people general guidelines on how to improve their photographs. Yet, there is such a thing as being too general — ’keep working on your composition, framing and lighting,’ is the advice offered on my second picture above. This is not particularly useful.
Nor can it be made more useful because Keegan is not looking at your pictures. Keegan is looking at data that describes your image and relating that to data describing images from those human photographers who define Keegan’s ‘understanding’ of what constitutes a good image. The end result is that the programme privileges a very particular kind of image that can easily be seen by clicking on the Keegan Hall of Fame. These images are overwhelmingly portraits — particularly head shots — wildlife pictures and pictures of flowers. That’s it. They are generally pleasant enough but that’s about it (once more, in my opinion). The specific choices of the developers — particularly the obsession with background blur — and the technical limitations of the programme determine this outcome and perhaps as the technology develops it might become more flexible, or more sophisticated in its ability to interpret the data. But it has a long way to go judging from the Hall of Fame.
The problem with Keegan and similar programmes is that for less experienced photographers Keegan’s interpretation of what constitutes a good photograph could become definitive for them. The end result, as with so much else on the internet, is stifling conformity and a reluctance to experiment and grow, to find a distinctive and personal photographic vision.
It’s still great fun to play with, though.