def initialize(self, opt):
self.opt = opt
self.root = opt.dataroot
split_file = os.path.join(self.root, 'dataset_' + opt.phase + '.txt')
self.A_paths = numpy.loadtxt(split_file,
dtype=str,
delimiter=' ',
skiprows=3,
usecols=(0))
self.A_paths = [os.path.join(self.root, path) for path in self.A_paths]
self.A_poses = numpy.loadtxt(split_file,
dtype=float,
delimiter=' ',
skiprows=3,
usecols=(1, 2, 3, 4, 5, 6, 7))
self.mean_image_depth = numpy.load(
os.path.join(self.root, 'mean_image_depth.npy'))
if opt.model == "poselstm":
self.mean_image = None
print("mean image subtraction is deactivated")
self.A_size = len(self.A_paths)
self.transform = get_posenet_transform(opt, self.mean_image_depth)
def initialize(self, opt):
self.opt = opt
self.root = opt.dataroot
split_file = os.path.join(self.root, 'dataset_' + opt.phase + '.txt')
self.A_paths = numpy.loadtxt(split_file,
dtype=str,
delimiter=' ',
skiprows=3,
usecols=(0))
self.A_paths = [os.path.join(self.root, path) for path in self.A_paths]
self.A_poses = numpy.loadtxt(split_file,
dtype=float,
delimiter=' ',
skiprows=3,
usecols=(1, 2, 3, 4, 5, 6, 7))
self.mean_image = numpy.load(os.path.join(self.root, 'mean_image.npy'))
self.A_size = len(self.A_paths)
self.transform = get_posenet_transform(opt, self.mean_image)
def initialize(self, opt):
self.opt = opt
self.root = opt.dataroot
split_file = os.path.join(self.root, 'dataset_' + opt.phase + '.txt')
self.A_paths = numpy.loadtxt(split_file,
dtype=str,
delimiter=' ',
skiprows=3,
usecols=(0))
self.A_paths = [os.path.join(self.root, path) for path in self.A_paths]
self.A_poses = numpy.loadtxt(split_file,
dtype=float,
delimiter=' ',
skiprows=3,
usecols=(1, 2, 3, 4, 5, 6, 7))
# if opt.isTrain:
# # scale values of location to defined range
# self.A_poses[:, :3], opt.position_range = scale(self.A_poses[:, :3], self.opt.scale_range)
# # TODO find a better way to store position_range
# file_name = os.path.join(self.opt.checkpoints_dir, self.opt.name, 'opt_'+self.opt.phase+'.txt')
# with open(file_name, 'at') as opt_file:
# opt_file.write('position_range: {}, {}\n'.format(opt.position_range[0], opt.position_range[1]))
# else:
# # read the position_range used for training from opt_train.txt
# path_train_file = os.path.join(opt.checkpoints_dir, opt.name, 'opt_train.txt')
# with open(path_train_file, 'rt') as ftrain:
# for line in ftrain:
# l = line.split(':')
# if 'position_range' == l[0]:
# opt.position_range = tuple(map(float, l[1].split(',')))
if opt.model == "poselstm":
self.mean_image = None
print("mean image subtraction is deactivated")
else:
self.mean_image = numpy.load(
os.path.join(self.root, 'mean_image.npy'))
self.A_size = len(self.A_paths)
self.transform = get_posenet_transform(opt, self.mean_image)