In yesterday post I pointed out the ongoing transformation in digital photography with an increased shift towards bits processing. Hence, it is not surprising that AI, Artificial Intelligence, is playing and will be playing a growing role.
AI is already at work at the shooting time, coordinating exposure and focussing, based on the “understanding” of the situation (is it a landscape, a portrait, …), and it is used by post processing applications, e.g. to detect foliage to make it greener or to replace some parts in an image with something that makes sense with the rest of the image (smart fill).
Now I run into the Spark a RenAIssance photo challenge launched by Huawei.
Huawei is calling for submission of digital images that will be evaluated by AI to judge how good they are (there will also be a professional photographer judging the photos submitted).
Judging a photo is a very subjective activity. Sure, there are some objective parameters like focus, white balance… but even those are very subjective. An out of focus image can provide a much better feeling, a stronger emotion, than a perfectly in-focus image.
Huawei is betting on AI to provide an evaluation on the artistic value of an image that in the end is what makes an image stand out. It will be interesting to compare the ranking resulting from the professional photographer with the one resulting from AI. I would not expect these to the the same (as I would not expect two professional photographers ending up with the same ranking) but I guess there should be some global consistency IF the AI has acquired a sensitivity on what is “beautiful” comparing with the one of (professional) photographers.
This may be the case since the AI is being trained on images that have been evaluated and ranked by professional photographers in many other challenges.
For me, Huawei challenge is clearly a marketing stunt, nevertheless it is an indicator that AI can play roles that would have been unthinkable in the past decade. Google at the end of 2017 introduced an AI based functionality to evaluate photos on structural and aesthetic parameters and more research results are being published in this area.