When we last featured JLaservideo’s E.D.I.T.H glasses, they used a notebook computer in a backpack, a webcam, and a see-through lens to display information about the world around them using image recognition techniques. Although interesting, the device was bulky enough that you wouldn’t want to lug it around with you everywhere.
In this second iteration, the unit has been reduced to a size that actually fits on the frame itself, using a Raspberry Pi Zero W and camera to take images. While Jlaser’ originally tried to run OpenCV on the Raspberry Pi, this proved problematic, and he decided to instead offload image processing to Google Images. This allows the setup to simply upload pictures to Dropbox, then a rather involved procedure is done to get Google Images to do the search, as no API is available. From there, data on what it sees is sent back to the glasses using a chat program. While the procedure is a bit kludged together, you have to admire how he was able to get this task done when normal programming techniques didn’t quite work.
Other notable points of the build include ripping apart a small LiPo charger to provide power, as well as implementing a modified HDMI heads-up display to project images on a section of the glasses itself. Be sure to check it out in the video below!
Go to Source
Author: Jeremy S. Cook