VISION: Visually Impaired Support Through Intelligent Object Narration

Visually impaired individuals often struggle to identify clothing and make independent shopping decisions. Existing object recognition systems provide only basic identification and lack detailed descriptions like color, pattern, or style. As a result, they rely on others to choose outfits or shop for new clothes. Our project helps bridge this gap by capturing an image of an apparel item, describing its key features in detail, and even suggesting similar options available online with their prices. This enables visually impaired users to explore and compare clothing independently, making fashion more accessible to them. We use deep learning models for apparel recognition and natural language processing (NLP) to generate accurate descriptions. A key feature of our system is Beam Search, which improves the accuracy of generated captions by selecting the most relevant descriptions. Additionally, APIs are integrated to fetch online comparisons, and a voice- based interface ensures easy interaction, making shopping more accessible and independent.