training format
Opened this issue · 2 comments
hi，i change the training code by write my own function
the landmakr format is x1,x2,,,,x68 y1,y2,,,,,y68.
after 5 stage my training error from 1.2--->0.05 with normalized
but when i test the result with my own model, only in the first frame
has a correct landmark ,and soon all landmarks disappeared..
what's my problem....
add test result and error
hi, i met the same problem.
when tested on each single image ,the landmarks is right. when tested on video,landmarks on the first frame were ok, while landmarks on subsequent frames are wrong and disappeared quickly.
i find that the landmark detection on the first frame is based on box by the face detector while those on on subsequent frames are based on boundingboxes of landmarks on the previous frame. the face boxes or the initial shapes are quite different, which leads different landmarks.
so have you solved the problems, and how?
many thx.