Neural Network Zaps You To Take Better Photographs

It’s ridiculously easy to take a bad photograph. Your brain is a far better Photoshop than Photoshop, and the amount of editing it does on the scenes your eyes capture often results in marked and disappointing differences between what you saw and what you shot.

Taking your brain out of the photography loop is the goal of [Peter Buczkowski]’s “prosthetic photographer.” The idea is to use a neural network to constantly analyze a scene until maximal aesthetic value is achieved, at which point the user unconsciously takes the photograph.

But the human-computer interface is the interesting bit — the device uses a transcutaneous electrical nerve stimulator (TENS) wired to electrodes in the handgrip to involuntarily contract the user’s finger muscles and squeeze the trigger. (Editor’s Note: This project is about as sci-fi as it gets — the computer brain is pulling the strings of the meat puppet. Whoah.)

Meanwhile, back in reality, it’s not too strange a project. A Raspberry Pi watches the scene through a Pi Cam and uses a TensorFlow neural net trained against a set of high-quality photos to determine when to trip the shutter. The video below shows it in action, and [Peter]’s blog has some of the photos taken with it.

We’re not sure this is exactly the next “must have” camera accessory, and it probably won’t help with snapshots and selfies, but it’s an interesting take on the human-device interface. And if you’re thinking about the possibilities of a neural net inside your camera to prompt you when to take a picture, you might want to check out our primer on TensorFlow to get started.

Thanks to [Peter]’s colleague [Julian], who tipped us off on this one.

38 thoughts on “Neural Network Zaps You To Take Better Photographs

  1. Spin-off thought; could you use electrical signals like this to stabilize the camera, effectively using your hand as a camera gimbal? It’d probably feel rather strange, but would actually be quite handy, and would be less cumbersome than a standard gimbal

    1. That’s what I thought this was doing. But apparently this is basically a computer controlled shutter. If only they made cheap little things that can plug into cameras and trigger them with a pure electrical signal….

  2. e·lec·tro·cute (ĭ-lĕk′trə-kyo͞ot′)
    tr.v. e·lec·tro·cut·ed, e·lec·tro·cut·ing, e·lec·tro·cutes
    1. To kill with electricity: a worker who was electrocuted by a high-tension wire.
    2. To execute (a person sentenced to death) by means of electricity.

  3. How is aesthetically pleasing defined? Rule of 3rds? Composition? Color? Subject? Lighting? Focus?
    It says the network is trained with high-quality photos, but what if you fed it experimental photography instead?

    I guess the point I’m getting at is that the artistic consensus of what is ‘aesthetically pleasing’ isn’t constant. Soup cans and urinals aside, I think we can all agree some photography or art that looks good to one person may not look good to another.

    1. Thats right. This particular example was trained with one particular dataset. (CUHKPQ which consists of 17,613 images. They were obtained from a variety of online communities and are divided into seven semantic categories. They were also labeled in categories of high- and low-quality pictures. A photographic community has done this by hand.) So each dataset results in a different overall aesthetic.
      This one for example especially “liked” the color blue, complete absence of color or stacked objects.
      There are many factors to what the particular algorithm classifies as aesthetically pleasing. The dataset, the neural network and also the learning time are all part of the result. Changing one can create different aesthetics. In the future, we may see different coexist and act as artists do. Some people might enjoy the aesthetics of algorithm X
      whereas others prefer the approach of algorithm Y.

  4. I’m wondering what was used to supply the TENS power. I’ve not seen any units that are both affordable AND easily controllable/triggerable from an outside source. It’s also not the sort of thing anybody should be DIYing as “good enough”, this is a device DESIGNED to pass a current through human muscle tissue. Not something you want to screw up.

    1. A commercial TENS unit was used 10€ on ebay(http://tenswelt.de/products/tns-sm-2-mf-tens-reizstromgeraet-mit-burst-und-modulation). Only the electordes are now stationary so one has not to deal with those messy and sticky stickers. It is powered with a 9v battery. One can adjust the mA till control over the contraction is lost. After all the testing I am still feeling fine :) Nothing else as using it as it was designed for. Except the computer decides when the signal is send.

  5. So much for creativity, when every photo looks like something from the training set.

    I’d get more use out of computer assistance at taking a group photo when everyone is smiling and has their eyes open. (Might have to do it by auto-combining multiple frames.)

    1. Sounds good in fact for such “classic, highly demanded type of photo, with high standard, and no artist touch” :p
      But I do think this can wipe out all imperfections we can see in brain-crafted photos where no one would have take this picture!

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.