Fig. 5: Demonstration of 4D-ONIX on experimental data. Four timestamps are shown, illustrating different stages of the additive manufacturing process.
From: 4D-ONIX for reconstructing 3D movies from sparse X-ray projections via deep learning
The three projections are shown in rows 1–3, corresponding to 0°, 27°, and 54°, respectively. Note that the scales of the horizontal and vertical directions differ, as we resized the image for faster computation and improved visualization. The top view and side view of the 3D ground truth are shown in rows 4–5. The top and side views of the 4D-ONIX reconstructions are shown in rows 6–7. The blue boxes mark the remelting regions in both the top and side views, while the red circles indicate an example area that poses challenges for the reconstruction algorithm. The bottom two rows show the Mean Squared Error (MSE) and Dissimilarity Structure Similarity Index Metric (DSSIM) between the 4D-ONIX reconstruction and the ground truth for each corresponding timestamp.
