A GPU based Deep Convolution Neural Network System to Object, Text and Colour for Android

Ritvij Saxena, B.S. Sanjay and S. Lavanya

The Existing aids and equipments for the visually impaired, only aid them to an extent where they can survive the world with ease, but, doesn’t provide sight or help them see what a normal person can see. With an Android application which helps detect objects, recognize text and identify colour, it would help them see the world to an extant more than they can by helping them identifying an object or the colour of the oject or any normal text present in them. We have come up with a prototype of an android application, which scans for the image dynamically and recognize its content based on the module the user is using and retrieve the type of object, or colour or text. The application software is trained with massive real-time dataset using Deep Convolutional Neural Network and other Machine learning Algorithms and APIs, such as tensorflow and OpenCV. When an object is scanned, it matches the object from the dataset using the training algorithm and then converts the result to text and later to speech. Since the use of smartphones is extant, it is simple to implify this model and hence makes it more feasible.

Volume 11 | 01-Special Issue

Pages: 241-247