NVIDIA CORP - 2021-06-1715 Jul 2021 18:28
NEURAL NETWORK BASED FACIAL ANALYSIS USING FACIAL LANDMARKS AND ASSOCIATED CONFIDENCE VALUES
Systems and methods for more accurate and robust determination of subject characteristics from an image of the subject. One or more machine learning models receive as input an image of a subject, and output both facial landmarks and associated confidence values. Confidence values represent the degrees to which portions of the subject's face corresponding to those landmarks are occluded, i.e., the amount of uncertainty in the position of each landmark location. These landmark points and their associated confidence values, and/or associated information, may then be input to another set of one or more machine learning models which may output any facial analysis quantity or quantities, such as the subject's gaze direction, head pose, drowsiness state, cognitive load, or distraction state.
[0003] Conventional gaze determination systems are not without their drawbacks, however. Overall performance and robustness of such systems remains limited, especially in edge cases or extreme situations such as large variations in head pose or partial occlusion of faces or eyes. FIG. 1 illustrates examples of such extreme situations. More specifically, FIG. 1 shows situations in which the subject's face is partially or mostly occluded (left), in this case by a respiratory mask, and in which the subject has turned her head far to one direction (right). In both examples, much of the subject's face cannot be seen (in the leftmost figure, the entirety of the subject's face below his eyes; in the rightmost figure, the entire right side of the subject's face), meaning much information on the subjects' faces is unavailable for sensing. This leaves limited information available for CNNs to determine face pose, resulting in often inaccurate results.
https://worldwide.espacenet.com/publicationDetails/description?CC=US&NR=2021182625A1&KC=A1&FT=D&ND=3&date=20210617&DB=&locale=en_EP
https://www.freepatentsonline.com/20210182625.pdf