Advances in the medical Image processing field anticipate the need to research and evaluate various applications in medical field. Biomedical is one such field that is considered most prominent in progress of diseases. Osteoporosis is the disease which inhibit the risk of fractures in human body due to weakening of bone. The bone becomes porous and weak which results in wear and tear of tissues resulting in fracture risk. In the present scenario Bone mass density (BMD) is well-known and acceptable standard by WHO to diagnose osteoporosis disease. In BMD scan the energy emitted by the x-ray beams is being passed through bone region and is absorbed and other part of the bone is not absorbed. So denser the bones mean it has good mineral content which absorbs more energy and less dense which absorbs less energy. The energy absorbed per pixel is measured in g/cm by converting. The density of bone is calculated with each pixel is based on the number of pixels in particular area. But still, it is questionable by many researchers. So motivated by this we have come up with custom x-ray images database for researchers to carry rigorous analysis in medical domain and forecast the findings by examining the accuracy of medical x-ray images results. our custom database of x-ray images has groups of Spine, Knee, Hand, Femur, Leg, Shoulder bones with details of Indian patients with unique patient ID, age, gender, diagnosis, image type. This Data can be used by young research scholars to work extensively in the osteoporosis bone disease and forecast promising results for betterment of medical society. In this study, we propose a Vision Transformer (ViT) model for classification of osteoporosis using X-ray images. The ViT model utilizes a self-attention mechanism to capture larger dependencies between image patches and has shown good results for various computer vision tasks. We trained and evaluated the model on a dataset of X-ray images, with images for each class normal and osteoporosis. The optimized hyperparameters were the RectifiedAdam optimizer with a learning rate of 0.0001, 500 epochs, and a mini-batch size of 16. The model achieved an overall accuracy of 88%. The proposed ViT classification model shows potential as a reliable and efficient tool for osteoporosis diagnosis, which could aid in early detection and treatment of the disease.