python - Trying to use dlib to measure mouth perimeter in a video and getting noisy results -


i have video of patient doing cognitive tasks. goal take each frame of video, face detection , landmark mouth, calculate area bounded mouth landmarks. used dlib , python api this.

i ran 2 problems. first, patient in hospital bed , camera view angled upward, looking @ chin instead of directly @ face. face isn't being detected in decent number of frames, , if there no face detected, algorithm doesn't try landmarking there no mouth perimeter points. wondering if there way can improve face detection (maybe training object detection on few frames of patient?)

the second problem frame frame, mouth landmarks can vary pretty significantly. hoping @ end of show slow motion word being spoken , mouth perimeter smoothly increasing , decreasing mouth opened , closed. but, result quite noisy bit of variation.

i'm method/platform agnostic. if knows of better, more accurate or robust way this, maybe matlab or opencv, open chasing lead. guidance helpful.

thanks, everyone.


Comments

Popular posts from this blog

angular - Ionic slides - dynamically add slides before and after -

minify - Minimizing css files -

Add a dynamic header in angular 2 http provider -