The Descriptive Camera Turns Images Into A Text Description

We've seen camera innovations in the realm of increases in aperture, megapixel count, and improvements in the backside illuminated sensor.  But Matt Richardson has taken this in a somewhat different approach, creating a device that outputs text describing the image it can see.

The camera hardware, created as part of a class for New York University's Interactive Telecommunications Program, is a USB webcam, a shutter button, an ethernet connection and a small thermal printer for outputting the results.  But the software is where this primarily gets interesting.  Once the shutter button is pressed the image is sent off for human analysis via Amazon's Mechanical Turk API.  From there the human will interpret the photo and write a description, returning it back to the camera for printing, even leaving a polaroid-esque photo outline.

It's a rather fascinating technological premise; but it does come at a price and limitation in its current early state.  The Amazon Human Intelligence Task (HIT for short) sets you back about $1.25 for each image, and the camera runs from an external 5-volt power source.  But he is continuing his work: forming an "accomplice mode" which allows the camera to send links to the image via instant messaging platforms for your contacts to interpret, dramatically reducing cost, modifying the device to run off self-contained batteries, and taking advantage of wireless data.

We've got to admit we rather enjoy the idea of having a camera that allows us to look at a scene through someone else's eyes.  Talk about 'every picture is worth a thousand words.'

Source: Matt Richardson

Jason England