Skip to main content
. 2020 Nov 28;19:100144. doi: 10.1016/j.smhl.2020.100144

Table 1.

Quantity of images resulting from successive face-related feature detections over the FFHQ dataset (70,000 images) and exploited for generating the MaskedFace-Net dataset (137,016 images) through mask-to-face mapping scenarios.

Results of applied detection techniques
Considered features Face Facial landmarks
Targeted mask-to-face mapping Correct Incorrect
Detection rate (over the FFHQ dataset) 99.75% 95.99%
Retained images 69,823 67,193 69,823