Skip to main content
Fig. 2 | BMC Oral Health

Fig. 2

From: Detecting representative characteristics of different genders using intraoral photographs: a deep learning model with interpretation of gradient-weighted class activation mapping

Fig. 2

Representative Gaussian heatmaps demonstrating interested regions of the R-net model which were marked with red color when dealing with uncropped and cropped images. Figure a–d showed that the R-net model paid more attention to (a) lip, (b)areas covered by soft tissue around maxillary teeth, (c) areas covered by soft tissue around lower posterior teeth, and (d) areas covered by soft tissue around lower anterior teeth when dealing with uncropped images. Figure e–i showed that the R-net model paid more attention to (e) upper anterior teeth, (f) lower anterior teeth, (g) marginal gingiva of lower anterior teeth, (h) occlusal region of posterior teeth, and (i) occlusal region of anterior teeth

Back to article page

BMC Oral Health

ISSN: 1472-6831

Contact us