Table 1 Comparison of model accuracy for different types of training data: real images, artificial noiseless images, and artificial images with noise added as a post-processing filter.

From: CT-based data generation for foreign object detection on a single X-ray projection

Training data, \(S_{train} = 21\)

Accuracy score A

Real data

\(91.1\% \pm 0.7\%\)

Noiseless artificial

\(91.8\% \pm 1.3\%\)

Noisy artificial

\(91.8\% \pm 0.4\% \)

Back to article page