This study uses AI-based object detection models Improved deNoising anchOr boxes (DINO) and You Only Look Once (YOLO) to detect and classify bone tumors on full-field radiographs. A five-fold cross-validation system was used to train and evaluate the object detection models. All cases were divided into five subsets then one subset was used as a test set while the other four subsets were used for training and validation (20% and 80% respectively). The result was that the DINO model was 85.7% accurate compared to the 80.1% accuracy of the YOLO model, with both being compared based on generated bounding boxes with the highest generated confidence scores. The generated data was cross checked by three orthopedic oncologists and three general orthopedic surgeons, with the DINO model performing better than the general orthopedic surgeons and performing comparably to the oncologists. The DINO model was also able to identify malignant tumors 76.8% of the time for problematic cases that yielded incorrect classifications from two or more doctors, with the malignancy/benign status being determined by pathological diagnoses through biopsy. Out of the six doctors, three general and three specialized oncologists, the DINO model tied second best with a specialist for accuracy, second best for sensitivity, fifth best for specificity, fourth best (better than all general doctors) for precision, and second highest for F-measure. DINO works due to its nature as a Transformer model, meaning it is able to analyze both local and global image features.