good day all, using google's new face api (link here) works improved facedetection. 1 of things list of landmark objects, each of has x , y coordinate use.
with these coordinates, trying figure out center of picture, figuring out correct numbers proving difficult.
here know far:
1) unlike in old ways, no longer 1000, 1000 -1000, -1000
2) the coordinates returned in float format , are, "...the (x, y) position of landmark (0, 0) upper-left corner of image. point guaranteed within bounds of image." (source)
3) when print out coordinates in log, numbers don't seem match screen size in pixels (1440w,2368h). examples of positions getting when print them out are:
- 464.90558,1112.7573
- -19.159714,218.88104
- 28.383072,196.1712
- -130.06908,1071.8779
which makes no sense since don't understand how coordinates can negative if top left (0,0).
so question is, how these coordinates being determined? respect size of screen somehow? being converted using dp in way? have static cap?
thanks all,
pgmac
coordinates pixel coordinates of image.
these coordinates can negative if face extends beyond top and/or left edges of image. though head may not entirely within photo, face detector estimate bounding box of face beyond image bounds based upon visible.
Comments
Post a Comment