Environment Descriptor for the Visually Impaired
Ankush Aniket Mishra, C. Madhurima, S Gautham, Jishin James, D. Annapurna · 2018
The ability to provide a system for automatically describing images to the users, especially the blind is a matter which can improve their lives significantly, if made possible to a user a friendly way. We present, a way of achieving the task of producing natural language descriptions of images through the use of recent deep architectures in neural networks, specifically based on recurrent and convolution architectures. A generative model for this is created and the accuracy for the MSCOCO Captioning dataset is presented along with a software architecture of how this can be extended as a standalone application for the blind.