This repository contains the code used in the paper "X-ray Image Generation As A Method Of Performance Prediction For Real-Time Inspection: A Case Study". It can be used for data generation, training and testing DCNNs, and performance evaluation using POD curves.
calibration.py was used to extract noise model parameters from the series of flatfield images.
real_data_process.py was used to make experimental datasets from raw data. The scipt also contains the code for computing dual-energy contrast used in POD curves.
noise_test.py was used to generate noisy datasets corresponding to different values of exposure time.
chichen.py was used to train and test DCNNs.
visualize_pod.py was used to plot POD curves, compare performance on real and generated data, and connect the performance estimate with exposure time.