def set_dataset(self):
"""properly handle multiple dataset situation
"""
test_fpath = os.path.join(os.path.dirname(os.path.dirname(__file__)),
"splits", self.opt.split, "test_files.txt")
self.val_filenames = readlines(test_fpath)
val_dataset = KittiDataset(
self.opt.data_path,
self.opt.val_gt_path,
self.val_filenames,
self.opt.height,
self.opt.width,
crph=self.opt.crph,
crpw=self.opt.crpw,
is_train=False,
predang_path=self.opt.predang_path,
semanticspred_path=self.opt.semanticspred_path,
threeinput=self.opt.threeinput)
self.val_loader = DataLoader(val_dataset,
batch_size=1,
shuffle=False,
num_workers=self.opt.num_workers,
pin_memory=True,
drop_last=True)
self.val_num = val_dataset.__len__()
mapping = readlines(
os.path.join(self.opt.trainmapping_fold, 'training_mapping.txt'))
self.indmapping = self.get_indmapping(mapping)
def set_dataset(self):
"""properly handle multiple dataset situation
"""
fpath = os.path.join(os.getcwd(), "splits", self.opt.split,
"{}_files.txt")
val_filenames = readlines(fpath.format("train"))
val_dataset = KittiDataset(
self.opt.data_path,
self.opt.gt_path,
val_filenames,
self.opt.height,
self.opt.width,
crph=self.opt.crph,
crpw=self.opt.crpw,
is_train=False,
semanticspred_path=self.opt.semanticspred_path)
self.val_loader = DataLoader(val_dataset,
self.opt.batch_size,
shuffle=True,
num_workers=0,
pin_memory=True,
drop_last=True)
self.val_num = val_dataset.__len__()
def set_dataset(self):
"""properly handle multiple dataset situation
"""
self.valdatasets = dict()
test_fpath = os.path.join(os.path.dirname(os.path.dirname(__file__)),
"splits", self.opt.split, "test_files.txt")
val_filenames = readlines(test_fpath)
val_dataset_rawLidar = KittiDataset(self.opt.data_path,
self.opt.gt_path_rawLidar,
val_filenames,
self.opt.height,
self.opt.width,
crph=self.opt.crph,
crpw=self.opt.crpw,
is_train=False)
val_dataset_filterLidar = KittiDataset(self.opt.data_path,
self.opt.gt_path_filterLidar,
val_filenames,
self.opt.height,
self.opt.width,
crph=self.opt.crph,
crpw=self.opt.crpw,
is_train=False)
val_dataset_semidense = KittiDataset(self.opt.data_path,
self.opt.gt_path_semidense,
val_filenames,
self.opt.height,
self.opt.width,
crph=self.opt.crph,
crpw=self.opt.crpw,
is_train=False)
self.valdatasets['rawLidar'] = DataLoader(
val_dataset_rawLidar,
self.opt.batch_size,
shuffle=False,
num_workers=self.opt.num_workers,
pin_memory=True,
drop_last=True)
self.valdatasets['filterLidar'] = DataLoader(
val_dataset_filterLidar,
self.opt.batch_size,
shuffle=False,
num_workers=self.opt.num_workers,
pin_memory=True,
drop_last=True)
self.valdatasets['semidense'] = DataLoader(
val_dataset_semidense,
self.opt.batch_size,
shuffle=False,
num_workers=self.opt.num_workers,
pin_memory=True,
drop_last=True)
def set_dataset(self):
"""properly handle multiple dataset situation
"""
fpath = os.path.join(os.path.dirname(os.path.dirname(__file__)),
"splits", self.opt.split, "{}_files.txt")
test_fpath = os.path.join(os.path.dirname(os.path.dirname(__file__)),
"splits", self.opt.split, "test_files.txt")
train_filenames = readlines(fpath.format("train"))
val_filenames = readlines(test_fpath)
train_dataset = KittiDataset(
self.opt.data_path,
self.opt.gt_path,
train_filenames,
self.opt.height,
self.opt.width,
crph=self.opt.crph,
crpw=self.opt.crpw,
is_train=True,
predang_path=self.opt.predang_path,
semanticspred_path=self.opt.semanticspred_path)
val_dataset = KittiDataset(
self.opt.data_path,
self.opt.val_gt_path,
val_filenames,
self.opt.height,
self.opt.width,
crph=self.opt.crph,
crpw=self.opt.crpw,
is_train=False,
predang_path=self.opt.predang_path,
semanticspred_path=self.opt.semanticspred_path)
self.train_loader = DataLoader(train_dataset,
self.opt.batch_size,
shuffle=True,
num_workers=self.opt.num_workers,
pin_memory=True,
drop_last=True)
self.val_loader = DataLoader(val_dataset,
self.opt.batch_size,
shuffle=True,
num_workers=self.opt.num_workers,
pin_memory=True,
drop_last=True)
def set_dataset(self):
"""properly handle multiple dataset situation
"""
fpath = os.path.join(os.getcwd(), "splits", self.opt.split, "{}_files.txt")
val_filenames = readlines(fpath.format("test"))
# val_filenames = readlines('/home/shengjie/Documents/Project_SemanticDepth/splits/eigen/test_files.txt')
val_dataset = KittiDataset(
self.opt.data_path, self.opt.gt_path, val_filenames, self.opt.height, self.opt.width,
crph=self.opt.crph, crpw=self.opt.crpw, is_train=False, instancepred_path=self.opt.instancepred_path
)
self.val_loader = DataLoader(
val_dataset, self.opt.batch_size, shuffle=False,
num_workers=0, pin_memory=True, drop_last=False)
self.val_num = val_dataset.__len__()
def set_dataset(self):
"""properly handle multiple dataset situation
"""
fpath = os.path.join(os.getcwd(), "splits", self.opt.split,
"{}_files.txt")
test_fpath = os.path.join(os.getcwd(), "splits", self.opt.split,
"test_files.txt")
train_filenames = readlines(fpath.format("train"))
val_filenames = readlines(test_fpath)
train_dataset = KittiDataset(self.opt.data_path,
self.opt.gt_path,
train_filenames,
self.opt.height,
self.opt.width,
crph=self.opt.crph,
crpw=self.opt.crpw,
is_train=True)
val_dataset = KittiDataset(self.opt.data_path,
self.opt.val_gt_path,
val_filenames,
self.opt.height,
self.opt.width,
crph=self.opt.crph,
crpw=self.opt.crpw,
is_train=False)
self.train_loader = DataLoader(train_dataset,
self.opt.batch_size,
shuffle=not self.opt.ban_shuffle,
num_workers=self.opt.num_workers,
pin_memory=True,
drop_last=True)
self.val_loader = DataLoader(val_dataset,
self.opt.batch_size,
shuffle=False,
num_workers=self.opt.num_workers,
pin_memory=True,
drop_last=True)
self.train_num = train_dataset.__len__()
self.val_num = val_dataset.__len__()
self.num_total_steps = self.train_num // self.opt.batch_size * self.opt.num_epochs