I need a Python model, built in an Anaconda environment, that can take a scanned fingerprint image in which only a fragment is visible and reconstruct the missing ridge patterns with high precision. The focus is strictly on reconstruction, not on downstream authentication or identification, so every design decision should optimize for ridge-level detail rather than classification accuracy. You will receive a folder of cropped and noise-added fingerprint fragments plus their full counterparts for validation. I am open to CNN-based inpainting, hybrid approach, AI and image processing—so long as the final output convincingly restores minutiae and passes a side-by-side comparison against the ground-truth scans. Please work in mainstream, well-documented libraries (PyTorch, TensorFlow or Keras, OpenCV, scikit-image) and keep the environment reproducible through a conda .yml file. Deliverables: • Fully commented Python scripts or notebooks • Trained model weights and instructions for inference on new scans • A concise README outlining setup, training, and evaluation steps • Quantitative report showing reconstruction quality (PSNR/SSIM or similar) on the supplied test set I expect code that runs end-to-end on a vanilla GPU machine and a short video or screenshot series that demonstrates the reconstruction process on at least three unseen partial prints.