Photo Captioning for Assisting Blind and Visually Impaired People Using Gated Attention
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
A people-centric workplace centers on individuals and their unique personal requirements. Translating theory into workplace design requires human-centered approaches that empathetically address employee needs. Voice-directed photo captioning and navigation systems can bring new horizons to visual data accessibility for visually impaired individuals, potentially improving their daily lives in ways most take for granted. We trained and deployed models on a single Graphics Processing Unit (GPU) of the Raspberry Pi 4B Single Board Computer (SBC). Blind and visually impaired safety and mobility are enhanced, especially in nighttime environments, with Non-Infrared (NoIR) cameras. In voice-aided image tagging for visually impaired users, we introduce a strategy combining Gated-Attention based models and YOLOv8 that leverages rich features from Deep Learning (DL) architectures for image description. Gated Attention models, implemented using Recurrent Neural Networks (RNN), ensure accuracy and consistency of output captions by combining the strengths of Gated Recurrent Units (GRUs) and attention mechanisms. Initially, an image is input into the YOLOv8 model to obtain a feature representation accommodating visual data. Upon receiving these characteristics, the Gated Attention-Based model generates sentences encapsulating the source image. A text-to-speech (TTS) application programming interface (API) converts written descriptions into spoken words. The audio output bridges visual and auditory realms, providing blind or visually impaired individuals with significant means of environmental understanding.
Fields of Study & Subjects
Multimodal Machine Learning Applications
Advanced Neural Network Applications
Assistive Technology in Communication and Mobility
Publication details
- Journal / source
- International Conference on Computing, Communication, Control and Cyber-Physical Systems (I5CPS)
- Pages
- 1–6
- Publication date
- 26 February 2026
- Language
- EN