Anushka Pote , Payal Malviya , Akanksha Pawar , Deepali Hajare , Varsha Pandagre

Innovations in machine learning and deep learning have got substantial progress in assistive devices made for individuals with visual impairments. This research paper presents an in-depth examination of existing ML and DL applications that aim to improve access to visual content for both the blind and visually impaired populations. Cutting-edge techniques such as neural networks for detailed feature extraction and enhanced AI technologies for generating descriptive image captions and detecting objects are carefully reviewed in this study. By using technologies like text-to-speech systems that convert the captions into audible descriptions for user interaction is greatly enhanced which makes content more accessible. This paper provides a thorough analysis of these advancements by assessing significant achievements, datasets applied and notable limitations present in current solutions. It offers comprehensive overview of field’s present status while identifying key gaps that restrict widespread adoption. Concluding with a discussion, this study suggests ways to improve assistive technology’s resilience, affordability and usability with a particular emphasis on inclusive design, the importance of diverse data sources and ongoing technological enhancement. Also it assesses how effectively various ML and DL frameworks adjust to diverse user needs and settings underscoring the versatility and potential of DL models for real-time processing of complex visual inputs in assistive applications.

https://doi.org/10.58357/jedcs.2024.v3.i4.2