In this paper, an innovative method is introduced to aid people with disabilities in navigating dimly lit surroundings by utilizing advanced machine learning and deep learning techniques. The application of Generative Adversarial Networks (GANs) for object recognition is investigated, demonstrating its superiority over conventional models such as Convolutional Neural Networks (CNN), Deep Residual Dense Networks (DRDN), Region-based CNN (RCNN), and High-Level Abstraction (HLA) algorithms. By identifying areas where existing research needs improvement, we suggest approaches that utilize complementary algorithms, such as Channel-wise Attention Block (CBAM), Squeeze-and-Excitation Networks (SE-Net), and Efficient Channel Attention Networks (ECA-Net), to enhance model performance. Our comprehensive method not only deals with object detection but also encompasses path prediction and obstacle avoidance, enabling independent navigation for individuals with disabilities. Through thorough mathematical analysis and comparison of our proposed models, we achieve significantly enhanced accuracy in low-light conditions, thereby improving the usability and effectiveness of assistive technologies for this demographic.