Affiliation:
1. Department of Computer Science and Engineering, Bhagwan Parshuram Institute of Technology, Delhi, India
Abstract
People with visual impairment (PVI) live a dependent life for doing their daily tasks. The most advanced assistive devices for PVIs currently available are either expensive or have poor performance or need extra hardware support. An integrated cognitive assistive tool for PVIs, HearUs, is implemented to handle path obstacle detection, product recognition, weather updates, reading documents or newspapers, automated dialing, and SOS service. It helps and directs the PVIs by giving them pertinent information about their surroundings. Different technologies have been used, like transfer learning, VGGNet 16 model for hurdle detection, barcode scanning for product identification, as well as text recognition and voice assistant capabilities. The Flickr8k dataset and a custom build dataset have been used for training the baseline model on around 8000 images. Furthermore, an experimental result shows that HearUs achieves 82% accuracy as compared to other existing applications.
Reference28 articles.
1. Development of an Android-based Learning Media Application for Visually Impaired Students
2. FingerReader2.0
3. AiSee
4. Text Reader for Visually Impaired Using Google Cloud Vision API.;P.Doshi;International Journal Of Innovative Research In Technology,2018
5. DeepDiary: Automatic caption generation for lifelogging image streams.;C.Fan;European Conference on Computer Vision,2016