Vision impairment, a leading disability worldwide, particularly prevalent in India, where the largest population of visually impaired individuals resides, prompts the need for innovative solutions. Thus, our study introduces a unique framework aimed at aiding the visually impaired by enabling object detection and recognition, empowering them to navigate their environment autonomously and enhance their situational awareness. The research employs transfer learning with the Single-Shot Detection (SSD) approach for identifying and categorizing objects, and subsequently, identifying human faces and currency notes if detected, using the Inception v3 model. To enable currency detection, the SSD detector is trained on a modified version of the PASCAL VOC 2007 dataset, with the inclusion of a new class. Moreover, separate Inception v3 models are trained for recognizing human faces and currency notes, ensuring flexibility and adaptability to user preferences within the framework. Ultimately, the framework's output can be presented audibly to visually impaired individuals. The Mean Accuracy and Precision (mAP) scores for the standalone SSD detector in detecting currency reached 67.8 percent, while the testing accuracy for recognizing individuals and currency notes using the Inception v3 model achieved 92.5 percent and 90.2 percent, respectively.

convolutional neural network, SSD, Inception v3, transfer learning