This document proposes a machine learning model to estimate bone age from pelvis X-ray images for individuals over 18 years of age. The model uses the Xception neural network architecture with transfer learning. The dataset contains 200 pelvis X-ray images of different age groups. The images are preprocessed using CLAHE and data augmentation. The model achieves a mean average error of 12.352 years for bone age estimation. This bone age estimation system could help determine the age of victims in forensic investigations when other identification documents are unavailable. The accuracy may be improved further by adding more X-ray images to the training dataset.