The prevalence of disorders affecting the vertebral column has increased recently as a result of injuries, sedentary lifestyle. Scoliosis and spondylolisthesis anomalies can start affecting kids early on and, if not addressed, can lead to terrible pain. Severe scoliosis might also result in heart and lung issues. Early diagnosis can therefore facilitate the application of treatments or interventions and stop the progression of the disease. Currently used diagnosis techniques rely on medical experts by visually inspecting radiographs. The observations are used by conventional AI-based diagnosis systems to carry out automated classification, enabling quick and simple diagnosis support tools. In this work, we leverage deep learning advancements in transfer learning to diagnose scoliosis and spondylolisthesis directly from X-ray pictures, without requiring any measurement. We gathered unprocessed data from authentic X-ray pictures showing spondylolisthesis, normal, and scoliosis. We also used data augmentation techniques to generalize the training process. Deep learning models, including DenseNet-121, and ResNet-152, were evaluated. For three-class classification, ResNet-152 performed significantly better than DenseNet-121 (94.42%), with a mean accuracy of 98.95%. The capacity to accurately diagnose the participants’ vertebral column disorders from routine X-ray pictures is demonstrated by the measured performance indicators. With minimal effort and errors, the current study offers a useful supporting technique to assist in early diagnosis and lowering the necessity for surgical procedures.