view article
Figure 5
Visualization of CNN prediction results using Grad-CAM. The red blocks in the image are the parts that make a significant contribution when CNN makes predictions. It is obtained as voxels 64 × 64 × 64 in size, but was divided into 64 2D images of 64 × 64 for plotting.

IUCrJ
Volume 11| Part 2| March 2024| Pages 182-189
ISSN: 2052-2525
https://doi.org/10.1107/S2052252524000770