Method and apparatus for generating landmark
Provided is a method of transforming a landmark including: receiving an input image including a facial image of a first person and a landmark corresponding to the facial image; estimating a transformation matrix corresponding to the landmark; and calculating an expression landmark and an identity landmark corresponding to the input image by using the transformation matrix.
1 . A method of transforming a landmark, the method comprising:
receiving an input image including a facial image of a first person and a facial landmark corresponding to the facial image;
estimating a transformation matrix corresponding to the facial landmark;
calculating, using the transformation matrix:
an expression landmark corresponding to the input image, wherein the expression landmark is related to a facial expression of the first person, and
an identity landmark corresponding to the input image, wherein the identity landmark is related to a unique identity of the first person; and
expressing the facial landmark as a sum of:
the expression landmark;
the identity landmark; and
an average landmark related to an average identity of human faces.
2 . The method of claim 1 , wherein estimating the transformation matrix comprises using a learning model trained to estimate a principal component analysis (PCA) transformation matrix from an arbitrary facial image and a landmark corresponding to the arbitrary facial image.
3 . The method of claim 2 , wherein the learning model:
classifies a plurality of landmarks into a plurality of semantic groups, and
outputs a PCA transformation coefficient corresponding to each of the plurality of semantic groups.
4 . The method of claim 3 , wherein calculating the expression landmark is performed using the transformation matrix and a PCA unit vector.
5 . The method of claim 1 , wherein the identity landmark comprises at least one of: texture information corresponding to the facial image, color information corresponding to the facial image, or shape information corresponding to the facial image.
6 . The method of claim 1 , wherein calculating the identity landmark comprises computing the expression landmark and the average landmark from the facial landmark.
7 . The method of claim 1 , wherein:
the expression landmark comprises expression information on at least one of: an eye of the facial image, a nose of the facial image, a mouth of the facial image, a facial contour of the facial image; or the facial image in entirety; and
the expression information corresponds to at least one of: position, angle, movement, inclination, or direction.
8 . A landmark generating apparatus comprising:
a memory; and
at least one processor, the at least one processor configured to:
receive an input image including a facial image of a first person and a facial landmark corresponding to the facial image;
estimate a transformation matrix corresponding to the facial landmark;
calculate an expression landmark and an identity landmark corresponding to the input image, wherein the identity landmark is related to a unique identity of the first person; and
express the facial landmark as a sum of:
the expression landmark;
the identity landmark; and
an average landmark related to an average identity of human faces.
9 . The landmark generating apparatus of claim 8 , wherein the at least one processor estimates the transformation matrix by using a learning model trained to estimate a principal component analysis (PCA) transformation matrix from an arbitrary facial image and a landmark corresponding to the arbitrary facial image.
10 . The landmark generating apparatus of claim 9 , wherein the learning model:
classifies a plurality of landmarks into a plurality of semantic groups, and
outputs a PCA transformation coefficient corresponding to each of the plurality of semantic groups.
11 . The landmark generating apparatus of claim 10 , wherein the at least one processor is configured to use the transformation matrix and a PCA unit vector to calculate an expression landmark corresponding to the facial image of the first person.
12 . The landmark generating apparatus of claim 8 , wherein:
the expression landmark comprises expression information on at least one of: an eye of the facial image, a nose of the facial image, a mouth of the facial image, a facial contour of the facial image; or the facial image in entirety;
the expression information corresponds to at least one of: position, angle, movement, inclination, or direction; and
the identity landmark comprises at least one of: texture information corresponding to the facial image, color information corresponding to the facial image, or shape information corresponding to the facial image.
13 . The landmark generating apparatus of claim 8 , wherein the at least one processor is configured to calculate the identity landmark by computing the expression landmark and the average landmark from the facial landmark.
14 . A non-transitory computer-readable recording medium having recorded thereon a program for performing a process comprising:
receiving an input image including a facial image of a first person and a facial landmark corresponding to the facial image;
estimating a transformation matrix corresponding to the facial landmark;
calculating, using the transformation matrix:
an expression landmark corresponding to the input image, wherein the expression landmark is related to a facial expression of the first person, and
an identity landmark corresponding to the input image, wherein the identity landmark is related to a unique identity of the first person; and
expressing the facial landmark as a sum of:
the expression landmark;
the identity landmark; and
an average landmark related to an average identity of human faces.
15 . The non-transitory computer-readable recording medium of claim 14 , wherein estimating the transformation matrix comprises using a learning model trained to estimate a principal component analysis (PCA) transformation matrix from an arbitrary facial image and a landmark corresponding to the arbitrary facial image.
16 . The non-transitory computer-readable recording medium of claim 15 , wherein the learning model:
classifies a plurality of landmarks into a plurality of semantic groups, and
outputs a PCA transformation coefficient corresponding to each of the plurality of semantic groups.
17 . The non-transitory computer-readable recording medium of claim 16 , wherein calculating the expression landmark is performed using the transformation matrix and a PCA unit vector.
18 . The non-transitory computer-readable recording medium of claim 14 , wherein the identity landmark comprises at least one of: texture information corresponding to the facial image, color information corresponding to the facial image, or shape information corresponding to the facial image.
19 . The non-transitory computer-readable recording medium of claim 14 , wherein calculating the identity landmark comprises computing the expression landmark and the average landmark from the facial landmark.
20 . The non-transitory computer-readable recording medium of claim 14 , wherein:
the expression landmark comprises expression information on at least one of: an eye of the facial image, a nose of the facial image, a mouth of the facial image, a facial contour of the facial image; or the facial image in entirety; and
the expression information corresponds to at least one of: position, angle, movement, inclination, or direction.