From 2e5a5ce00d0a85c8f1dd6e8790916e98b69a8fd3 Mon Sep 17 00:00:00 2001
From: Guillaume HEUSCH <guillaume.heusch@idiap.ch>
Date: Wed, 26 Feb 2020 10:13:42 +0100
Subject: [PATCH] [FaceCropAlign] inversed x,y coordinates when getting eyes
and mouth center, to be consistent with how annotations are computed and
sored
---
bob/pad/face/preprocessor/FaceCropAlign.py | 18 +++++++++++++-----
1 file changed, 13 insertions(+), 5 deletions(-)
diff --git a/bob/pad/face/preprocessor/FaceCropAlign.py b/bob/pad/face/preprocessor/FaceCropAlign.py
index 387e7681..a4b76bb6 100644
--- a/bob/pad/face/preprocessor/FaceCropAlign.py
+++ b/bob/pad/face/preprocessor/FaceCropAlign.py
@@ -102,7 +102,8 @@ def get_mouth_center(lm):
mouth_center_t = (lm[48, :] + lm[54, :]) / 2.0
- mouth_center = (int(mouth_center_t[1]), int(mouth_center_t[0]))
+ #mouth_center = (int(mouth_center_t[1]), int(mouth_center_t[0]))
+ mouth_center = (int(mouth_center_t[0]), int(mouth_center_t[1]))
return mouth_center
@@ -159,7 +160,8 @@ def get_eye_center(lm):
left_eye_t = (lm[36, :] + lm[39, :]) / 2.0
right_eye_t = (lm[42, :] + lm[45, :]) / 2.0
- eye_center = (int((left_eye_t[1]+right_eye_t[1])/2.0), int((left_eye_t[0]+right_eye_t[0])/2.0))
+ # eye_center = (int((left_eye_t[1]+right_eye_t[1])/2.0), int((left_eye_t[0]+right_eye_t[0])/2.0))
+ eye_center = (int((left_eye_t[0]+right_eye_t[0])/2.0), int((left_eye_t[1]+right_eye_t[1])/2.0))
return eye_center
@@ -304,6 +306,10 @@ def normalize_image_size_in_grayscale(image, annotations,
An image of the cropped face of the size (face_size, face_size).
"""
+
+ # WARNING : landmarks existing as annotations for BATL2 are (x,y) and not (y,x)
+ # As a consequence, when loading new landmarks, eyes and mouth center are wrong ...
+
if use_face_alignment:
@@ -323,7 +329,6 @@ def normalize_image_size_in_grayscale(image, annotations,
elif alignment_type=='lightcnn': # This option overrides the facesize argument
# This is the size of the image that this model expects
-
CROPPED_IMAGE_HEIGHT = 128
CROPPED_IMAGE_WIDTH = 128
@@ -345,6 +350,7 @@ def normalize_image_size_in_grayscale(image, annotations,
annotations['mouth_center']=mouth_center
+
light_cnn_face_cropper=bob.bio.face.preprocessor.FaceCrop(
cropped_image_size=(CROPPED_IMAGE_HEIGHT, CROPPED_IMAGE_WIDTH),
cropped_positions={'eye_center': EYE_CENTER_POS, 'mouth_center': MOUTH_CENTER_POS})
@@ -415,6 +421,7 @@ def normalize_image_size(image, annotations, face_size,
RGB 3D or gray-scale 2D.
"""
+
if len(image.shape) == 3:
if not (rgb_output_flag):
@@ -426,9 +433,10 @@ def normalize_image_size(image, annotations, face_size,
image = [image] # make gray-scale image an iterable
result = []
+
for image_channel in image: # for all color channels in the input image
-
+
cropped_face = normalize_image_size_in_grayscale(
image_channel, annotations, face_size, use_face_alignment,alignment_type=alignment_type)
@@ -621,7 +629,7 @@ class FaceCropAlign(Preprocessor):
An image of the cropped / aligned face, of the size:
(self.face_size, self.face_size), RGB 3D or gray-scale 2D.
"""
-
+
# sanity check:
if not self.rgb_output_flag and len(image.shape) != 2:
logger.warning("This image has 3 channels")
--
GitLab