We present a novel method for identifying traumatic maxillofacial fractures using convolutional neural networks with transfer learning (MFDS). A model for the categorization of future computed tomography (CT) scans as "fracture" or "noFracture" was developed by re-training a convolutional neural network previously trained on non-medical pictures using CT scans. There were a total of of 148 CT scans used to train the model (120 patients were identified as having a fracture, and 28 were labeled as having no fracture). There were a total of 30 patients included in the validation dataset utilized for statistical analysis (5 with "noFracture" and 25 with "fracture"). An additional 30 CT scans were utilized as the test dataset, including 25 "fracture" pictures and 5 "noFracture" images. Both a focus on individual slices and on grouped slices for patients was used in the tests. If the likelihood of a fracture in two successive slices was more than 0.99, the patient was considered to have a fracture. Patient data demonstrates that the model achieves an 80% rate of success in diagnosing maxillofacial fractures. Even while the MFDS model can't take the position of a radiologist, it can be a huge help in many ways: lowering the likelihood of mistakes, keeping patients safe by shortening the time it takes to get a diagnosis, and lightening the load of being hospitalized