Optimal

Live feed Camera

Visual feedback is acquired from a camera mounted on the users forehead.

Pre-processing Image

Video feed acquired is pre-processed with techniques like: histogram normalization.

Object detector

The pre-processed image is run through a pre-trained deep neural network to detect objects in an image.

Object classification

Objects detected are classified using a pre-trained deep neural network.

Object distance integration

Objects classified and detected are integrated with a distance calculation algorithm, to get their distance from the user.

Audio feedback

Object name and its distance are sent to the user via audio cue.

Our Goals

OPTImization & MAchine Learning (OPTIMAL) lab is developing wearable glasses to assist the visually impaired in interacting with the environment using deep neural networks. The system not only detects obstacles but also categorize objects and their relative distances from the user. The information will be fed to the user in the form of an audio feedback.