Open freebooterish opened 5 years ago
Hi,
There is a hard limit set in code in FaceLandmarkVidMulti for number of faces it tracks simultaneously (as it needs to keep some state for each face for tracking), you can actually change that hard limit by changing the following line in FaceLandmarkVidMulti.cpp:
int num_faces_max = 4;
The main difference is that FaceLandmarkImg treats each image separately while FaceLandmarkVidMulti first detects faces and then attempts to track them in a video sequence (with re-detection when the model fails).
Thanks, Tadas
Hi, Thanks your so amazing project, I found one simple question here. When I use FaceLandmarkVidMulti(with -wild and -multi_view 1), It can only detects 4~ faces per frame, BUT when I use FaceLandmarkImg (the same video file,using ffmpeg to extract exery frame, such as: ffmpeg -i test5.mp4 -r 1 -f image2 image-%3d.jpeg ) , the OP can detects 12~ faces per frame Could you please share me Why the two routines different? For detecting more faces, Should I use FaceLandmarkImg for every frame? the attachment are the two result CSVs, thanks!
result.zip