The most recent survey estimates that approximately 1.5 million visually impaired individuals live in India and struggle to carry out daily tasks. According to the most recent statistics, this population totals about 170 million around the globe and is growing at a 10% annual rate. Visually disabled people have a difficult time navigating new environments and knowing where a staircase is poses a significant challenge. Identifying the stairs is difficult due to various types of staircases, various viewing angles, and poor lighting conditions. Classifying them as up or down is even more difficult. For vision impaired people, however, finding out downstairs is even more crucial. The stair recognition algorithm should also be quick and accurate. This paper proposes a deep learning-based system for helping staircase detection and classify them as going up or down. A custom-made stairs dataset is used to test different models and the best model found is the fine-tuned VGG-19 deep learning model.