Object Detection for the Visually Impaired: A Systematic Literature Review

The development of object detection devices has become a significant investigative focus to enhance autonomy and quality of life for the visually impaired. In this context, technologies of computer vision are the predominant basis for evaluating and comparing the performance of several current object detection models that have the potential to be employed as the main component in assistive devices for the visually impaired. This inquiry presented experimental results and performance evaluations demonstrating the existence of approaches offering a good balance between detection accuracy and inference speed, a pivotal criterion in the development of object detection devices. However, significant challenges remain to be resolved, notably regarding the dependence of model performance on the utilized methodological approaches and the quality of the available datasets. Therefore, further steps are needed to overcome it. In this paper, future studies are proposed to focus on developing more complex and specialized model architectures to increase object detection accuracy and reduce inference time. Additionally, the diversification of datasets is an imperative concern to ensure broader coverage of possible usage situations. The integration of additional features (e.g. speech recognition and object distance estimation) is also proposed to improve the usability of object detection aids for the visually impaired. By taking these steps, it is expected to produce a more sophisticated, reliable, and better-suited object detection device based on the user's needs. Apart from direct benefits for the visually impaired groups, the contribution of this study is also expected to enrich the literature in the computer vision domain, particularly in the context of applying technology to improve human welfare in general.