def visualize_segmentation(input_img,
seg_arr,
n_classes,
class_number=2,
display=False,
output_file=None):
seg_img = predict.segmented_image_from_prediction(
seg_arr, n_classes=n_classes, input_shape=input_img.shape)
overlay_img = cv2.addWeighted(input_img, 0.7, seg_img, 0.3, 0)
# Reshaping the Lanes Class into binary array and Upscaling the image as input image
dummy_img = np.zeros(seg_arr.shape)
dummy_img += ((seg_arr[:, :] == class_number) * (255)).astype(
'uint8') # Class Number 2 belongs to Lanes
original_h, original_w = overlay_img.shape[0:2]
upscaled_img = cv2.resize(dummy_img,
(original_w, original_h)).astype('uint8')
upscaled_img_rgb = cv2.cvtColor(upscaled_img, cv2.COLOR_GRAY2RGB)
# Stack input and segmentation in one video
vis_img = np.vstack((
np.hstack((input_img, seg_img)),
np.hstack((
overlay_img,
upscaled_img_rgb)) # np.ones(overlay_img.shape,dtype=np.uint8)*128
))
return upscaled_img #vis_img #vis_img
def visualize_segmentation(input_img,
seg_arr,
n_classes,
class_number=2,
display=False,
output_file=None):
seg_img = predict.segmented_image_from_prediction(
seg_arr, n_classes=n_classes, input_shape=input_img.shape)
overlay_img = cv2.addWeighted(input_img, 0.7, seg_img, 0.3, 0)
# Reshaping the Lanes Class into binary array and Upscaling the image as input image
dummy_img = np.zeros(seg_arr.shape)
dummy_img += ((seg_arr[:, :] == class_number) * (255)).astype(
'uint8') # Class Number 2 belongs to Lanes
original_h, original_w = overlay_img.shape[0:2]
upscaled_img = cv2.resize(dummy_img,
(original_w, original_h)).astype('uint8')
upscaled_img_rgb = cv2.cvtColor(upscaled_img, cv2.COLOR_GRAY2RGB)
# centerline_arr = np.mean(upscaled_img, axis = 1)
# for c_r in range(len(centerline_arr)):
# cv2.circle(upscaled_img_rgb, (int(centerline_arr[c_r]), c_r), 5, (0, 0, 255), -1)
# for c_r in range(upscaled_img_rgb.shape[0]):
# nonzero_xy = upscaled_img_rgb[c_r][1:].nonzero()
# nonzero_x = np.array(nonzero_xy[0])
#
# centerline_arr = np.mean(nonzero_x)
#
# if np.isnan(centerline_arr)==0:
# # print centerline_arr #img[c_r][1:].nonzero(),
#
# cv2.circle(upscaled_img_rgb, (int(centerline_arr), c_r), 5, (0, 0, 255), -1)
# Stack input and segmentation in one video
vis_img = np.vstack((
np.hstack((input_img, seg_img)),
np.hstack((
overlay_img,
upscaled_img_rgb)) # np.ones(overlay_img.shape,dtype=np.uint8)*128
))
return upscaled_img_rgb #upscaled_img
def visualize_segmentation(self, seg_arr, display=False, output_file=None):
seg_img = predict.segmented_image_from_prediction(
seg_arr, self.n_classes, input_shape=self.image.shape)
# Reshaping the Lanes Class into binary array and Upscaling the image as input image
self.overlay_img = cv2.addWeighted(self.image, 0.7, seg_img, 0.3, 0)
dummy_img = np.zeros(seg_arr.shape)
dummy_img += ((seg_arr[:, :] == self.label_no) * (255)).astype('uint8')
original_h, original_w = self.overlay_img.shape[0:2]
self.pred_img = cv2.resize(dummy_img,
(original_w, original_h)).astype('uint8')
upscaled_img_rgb = cv2.cvtColor(self.pred_img, cv2.COLOR_GRAY2RGB)
# Stack input and segmentation in one video
vis_img = np.vstack((
np.hstack((self.image, seg_img)),
np.hstack((self.overlay_img, upscaled_img_rgb
)) # np.ones(overlay_img.shape,dtype=np.uint8)*128
))
def visualize_segmentation(
self
): #input_img, seg_arr, n_classes, display = False, output_file = None, class_number = 2
seg_img = predict.segmented_image_from_prediction(
self.seg_arr,
n_classes=self.model.n_classes,
input_shape=self.image.shape)
overlay_img = cv2.addWeighted(self.image, 0.7, seg_img, 0.3, 0)
# Reshaping the Lanes Class into binary array and Upscaling the image as input image
dummy_img = np.zeros(self.seg_arr.shape)
dummy_img += ((self.seg_arr[:, :] == self.class_number) *
(255)).astype('uint8') # Class Number 2 belongs to Lanes
original_h, original_w = overlay_img.shape[0:2]
upscaled_img = cv2.resize(dummy_img,
(original_w, original_h)).astype('uint8')
upscaled_img_rgb = cv2.cvtColor(upscaled_img, cv2.COLOR_GRAY2RGB)
# Stack input and segmentation in one video
vis_img = np.vstack((
np.hstack((self.image, seg_img)),
np.hstack((overlay_img, upscaled_img_rgb
)) #np.ones(overlay_img.shape,dtype=np.uint8)*128))
))
#try:
(trans, rot) = self.listener.lookupTransform('utm', 'base_link',
rospy.Time(0))
#except (tf.LookupException, tf.ConnectivityException, tf.ExtrapolationException):
#continue
robot_lat_utm = trans[1] # Latitude
robot_long_utm = trans[0] # Longitude
a = np.array((robot_lat_utm, robot_long_utm))
b = np.array((self.gt_lat_utm[1], self.gt_long_utm[1]))
self.dist_0 = np.linalg.norm(a - b)
#yaw_gt = math.atan2(gt_long_utm[1],gt_lat_utm[1])
for i in range(2, len(self.gt_lat_utm)):
b = np.array((self.gt_lat_utm[i], self.gt_long_utm[i]))
dist = np.linalg.norm(a - b)
if dist < self.dist_0:
self.dist_0 = dist
# #self.dist_0 = self.dist_0+1
print self.dist_0
font = cv2.FONT_HERSHEY_SIMPLEX
bottomLeftCornerOfText = (100, 70)
fontScale = 2
fontColor = (255, 0, 0)
lineType = 2
cv2.putText(vis_img, 'Lateral Offset:' + str(self.dist_0),
bottomLeftCornerOfText, font, fontScale, fontColor,
lineType)
# cv2.imshow('preview', vis_img)
#
# # Press Q on keyboard to exit
# if cv2.waitKey(25) & 0xFF == ord('q'):
# print('Q pressed, breaking')
return vis_img