If you’ve ever dabbled in photography, the first piece of advice you likely received was to take as many pictures as possible. Practice makes perfect after all, and there is a reason even seasoned photographers snap a lot of photos on their smartphones. But, just walking around taking out-of-focus pictures of every leaf you see isn’t enough, you also need to learn to get shots that are actually good.
To assist in that endeavor, Peter Buczkowski created the Prosthetic Photographer. This is a kind of camera augmentation that alerts you when it thinks you’ve got a nice composition lined up. The first noteworthy part of the system is how it notifies you when it’s time to take a picture.
Embedded in the handle of the Prosthetic Photographer is a pair of metal contacts, which give you a mild shock when you should snap a shot. That shock is delivered from a TENS (Transcutaneous Electrical Nerve Stimulation) unit. If you can manage to not drop the camera in pain and frustration, you can then push a button on the handle to trigger the camera release.
The real magic, however, is the system’s ability to know if a photo is going to be good. That all happens on a Raspberry Pi 3 that’s running the Inception model on Google’s TensorFlow neural network. The model is able to continuously scan the world through a Raspberry Pi Camera Module, and uses deep learning to recognize a potentially nice picture. As you can see, it seems to be pretty good at that.