I need a complete machine-learning pipeline that can look at medical images—specifically plain-film X-rays—and tell me whether each study is of the chest, abdomen, or an extremity. All input files will be standard hospital exports (mostly DICOM, occasionally PNG/JPEG), so the model must handle typical variations in resolution and contrast. What I’m after is a reproducible, well-documented solution: data preparation, augmentation, model architecture (a CNN in TensorFlow, Keras, or PyTorch is fine), training, and evaluation. Please include class-balanced splits, explain any preprocessing you apply, and show the metrics you achieve on an unseen validation set. Deliverables • Python code with clear comments for preprocessing, training, and inference • Trained model weights ready for deployment • A short report (or notebook) summarising accuracy, confusion matrix, and any tricks you used • Simple CLI or notebook that lets me drag-and-drop new chest, abdomen, or extremity X-rays and get the predicted class If you already have experience with radiology datasets or have tackled similar chest/abdomen/extremity X-ray projects, mention it—speed and reliability matter to me.