view article
Figure 3
Architecture of the CNN used for 2D XRD pattern classification. The model consists of three main components: (1) a convolutional backbone with four blocks for feature extraction, each containing multiple convolutional layers and max pooling, (2) fully connected layers with 4096 hidden neurons for further feature refinement, and (3) classification heads for 230-way and 7-way crystal system classification. The model is trained using a dataset split in a 7:2:1 ratio for training, validation and testing, optimized with the Adam optimizer and with a learning rate of 0.0001 over 100 epochs.

Journal logoJOURNAL OF
APPLIED
CRYSTALLOGRAPHY
ISSN: 1600-5767
Volume 59| Part 1| February 2026| Pages 206-224
https://doi.org/10.1107/S1600576726000099
Follow J. Appl. Cryst.
Sign up for e-alerts
E-alerts
Follow J. Appl. Cryst. on Twitter
Twitter
Follow us on facebook
Facebook
Sign up for RSS feeds
RSS