Most cameras today have settings that log the time, date, camera settings, and even location of any photos taken. Unfortunately though, no camera out there can automatically note what a picture actually shows ... until now, that is. Matt Richardson, a student in the Computational Cameras class at New York University‘s Interactive Telecommunications Program, recently created a quirky device called the "Descriptive Camera," which works like a regular camera, but instead of displaying images, it prints out a description of the photo's content in plain English.
Like any other camera, the Descriptive Camera takes photos when someone presses the shutter while pointing it at a subject. It's what happens after that button is pressed though that makes this camera unique. The recorded image is sent off to be viewed by a number of people on the internet, who create a crowdsourced description of the photo's content. This process typically takes around three to six minutes and is indicated by a yellow light labeled "Developing" on the camera lighting up. Once the camera receives the final description, a thermal printer dispenses it like a receipt from the grocer. This way, a picture of a set of drawers and a lamp can turn into a passage saying, "Looks like a cupboard which is ugly and old having name plates on it with a study lamp attached to it."
Richardson created a the camera as a project for New York University's Interactive Telecommunications Program. The prototype is comprised of a USB webcam and thermal printer attached to a BeagleBone Linux board, which controls the camera's functions and connects to the internet through an ethernet cable. "Developing" the description of an image requires it to be sent through Amazon's Mechanical Turk API, which submits the task to human workers on the internet, who come up with a short summary for US$1.25 each. The Mechanical Turk API also offers a ranking and approval system to encourage accurate results.
It may not be incredibly useful to have a camera that records text-only versions of pictures, but technology like this could certainly lead to better methods for automatically logging and sorting images. In the future, Richardson hopes to improve on the prototype by adding a battery and wireless capabilities to make it feel more like a regular digital camera.
Source: Matt Richardson