def get_colored_lidar(self, idx):
"""
Returns a point cloud with each pt containing not only the xyz (for now no reflectance)
but also the color of its projection onto the image.
Args:
idx ():
Returns:
colored_pts: (n, 3 + 3), where the 6 are xyz + rgb
"""
# get the lidar
# todo: call get_lidar instead
lidar_file = os.path.join(self.root_split_path, 'velodyne',
'%s.bin' % idx)
assert os.path.exists(lidar_file)
pts = np.fromfile(lidar_file, dtype=np.float32).reshape(-1, 4)
pts = pts[:, :3]
# get the image
img_file = os.path.join(self.root_split_path, 'image_2',
'%s.png' % idx)
assert os.path.exists(img_file)
img = np.array(io.imread(img_file), dtype=np.int32)
img_shape = img.shape # should be (heigth, width)
# get the calibration
calib_file = os.path.join(self.root_split_path, 'calib',
'%s.txt' % idx)
assert os.path.exists(calib_file)
calib = calibration.Calibration(calib_file)
# filter out points that are not within image fov
# todo: this migth already be done in other preprocesing, but might be later on
pts_rect = calib.lidar_to_rect(
pts[:, :3]) # get_fov_flag only works with rect coord sys
fov_flag = self.get_fov_flag(pts_rect, img_shape, calib)
pts_fov = pts[fov_flag]
# project pts onto image to get the point color
img_coords, _ = calib.lidar_to_img(pts_fov[:, :3])
img_coords = img_coords.astype(
np.int) # note that first col is the cols, second col is the rows
rows = img_coords[:, 1]
cols = img_coords[:, 0]
colors = img[rows, cols]
# todo: potential way to play with the semantics
colors = colors.astype(np.float32)
# colors /= 255 # normalize to [0, 1]
colors *= 0
# append each pt with its color
colored_pts = np.hstack([pts_fov, colors])
return colored_pts
def get_calib(self, idx):
calib_file = os.path.join(self.root_split_path, 'calib',
'%s.txt' % idx)
assert os.path.exists(calib_file)
return calibration.Calibration(calib_file)
