def __init__(self,
config,
input_transform=None,
target_transform=None,
augment_data=True,
elastic_distortion=False,
cache=False,
phase=DatasetPhase.Train):
if isinstance(phase, str):
phase = str2datasetphase_type(phase)
# Use cropped rooms for train/val
data_root = config.scannet_path
if phase not in [DatasetPhase.Train, DatasetPhase.TrainVal]:
self.CLIP_BOUND = self.TEST_CLIP_BOUND
data_paths = read_txt(
os.path.join(data_root, self.DATA_PATH_FILE[phase]))
logging.info('Loading {}: {}'.format(self.__class__.__name__,
self.DATA_PATH_FILE[phase]))
super().__init__(data_paths,
data_root=data_root,
input_transform=input_transform,
target_transform=target_transform,
ignore_label=config.ignore_label,
return_transformation=config.return_transformation,
augment_data=augment_data,
elastic_distortion=elastic_distortion,
config=config)
def load_extrinsics(extrinsics_file):
"""Load the camera extrinsics from a .txt file.
"""
lines = read_txt(extrinsics_file)
params = [float(x) for x in lines[0].split(' ')]
extrinsics_matrix = np.asarray(params).reshape([4, 4])
return extrinsics_matrix
def __init__(self,
config,
prevoxel_transform=None,
input_transform=None,
target_transform=None,
augment_data=True,
elastic_distortion=False,
cache=False,
phase=DatasetPhase.Train):
if isinstance(phase, str):
phase = str2datasetphase_type(phase)
if phase not in [DatasetPhase.Train, DatasetPhase.TrainVal]:
self.CLIP_BOUND = self.TEST_CLIP_BOUND
data_root = config.synthia_path
data_paths = read_txt(
osp.join('./splits/synthia4d', self.DATA_PATH_FILE[phase]))
data_paths = [d.split()[0] for d in data_paths]
logging.info('Loading {}: {}'.format(self.__class__.__name__,
self.DATA_PATH_FILE[phase]))
super().__init__(data_paths,
data_root=data_root,
input_transform=input_transform,
target_transform=target_transform,
ignore_label=config.ignore_label,
return_transformation=config.return_transformation,
augment_data=augment_data,
elastic_distortion=elastic_distortion,
config=config)
def __init__(self,
config,
prevoxel_transform=None,
input_transform=None,
target_transform=None,
cache=False,
augment_data=True,
elastic_distortion=False,
phase=DatasetPhase.Train):
if isinstance(phase, str):
phase = str2datasetphase_type(phase)
if phase not in [DatasetPhase.Train, DatasetPhase.TrainVal]:
self.CLIP_BOUND = self.TEST_CLIP_BOUND
data_root = config.data.stanford3d_path
if isinstance(self.DATA_PATH_FILE[phase], (list, tuple)):
data_paths = []
for split in self.DATA_PATH_FILE[phase]:
data_paths += read_txt(os.path.join(data_root, 'splits',
split))
else:
data_paths = read_txt(
os.path.join(data_root, 'splits', self.DATA_PATH_FILE[phase]))
if config.data.voxel_size:
self.VOXEL_SIZE = config.data.voxel_size
logging.info('voxel size: {}'.format(self.VOXEL_SIZE))
logging.info('Loading {} {}: {}'.format(self.__class__.__name__, phase,
self.DATA_PATH_FILE[phase]))
VoxelizationDataset.__init__(
self,
data_paths,
data_root=data_root,
prevoxel_transform=prevoxel_transform,
input_transform=input_transform,
target_transform=target_transform,
ignore_label=config.data.ignore_label,
return_transformation=config.data.return_transformation,
augment_data=augment_data,
elastic_distortion=elastic_distortion,
config=config)
def load_intrinsics(intrinsics_file):
"""Load the camera intrinsics from a intrinsics.txt file.
intrinsics.txt: a text file containing 4 values that represent (in this order) {focal length,
principal-point-x, principal-point-y, baseline (m) with the corresponding right
camera}
"""
lines = read_txt(intrinsics_file)
assert len(lines) == 7
intrinsics = {
'focal_length': float(lines[0]),
'pp_x': float(lines[2]),
'pp_y': float(lines[4]),
'baseline': float(lines[6]),
}
return intrinsics
def load_datafile(self, index):
filepath = self.data_root / self.data_paths[index]
scene_id = os.path.splitext(self.data_paths[index].split(
os.sep)[-1])[0]
scene_f = self.config.scannet_alignment_path % (scene_id, scene_id)
ptc = read_plyfile(filepath)
if os.path.isfile(scene_f):
alignment_txt = [
l for l in read_txt(scene_f)
if l.startswith('axisAlignment = ')
][0]
rot = np.array([float(x) for x in alignment_txt[16:].split()
]).reshape(4, 4)
xyz1 = np.hstack((ptc[:, :3], np.ones((ptc.shape[0], 1))))
ptc[:, :3] = (xyz1 @ rot.T)[:, :3]
return ptc, None, None
def __init__(self,
config,
prevoxel_transform=None,
input_transform=None,
target_transform=None,
augment_data=True,
elastic_distortion=False,
cache=False,
phase=DatasetPhase.Train):
if isinstance(phase, str):
phase = str2datasetphase_type(phase)
if phase not in [DatasetPhase.Train, DatasetPhase.TrainVal]:
self.CLIP_BOUND = self.TEST_CLIP_BOUND
data_root = config.synthia_path
data_paths = read_txt(osp.join(data_root, self.DATA_PATH_FILE[phase]))
data_paths = sorted([d.split()[0] for d in data_paths])
seq2files = defaultdict(list)
for f in data_paths:
seq_name = f.split(os.sep)[0]
seq2files[seq_name].append(f)
self.camera_path = config.synthia_camera_path
self.camera_intrinsic_file = config.synthia_camera_intrinsic_file
self.camera_extrinsics_file = config.synthia_camera_extrinsics_file
# Force sort file sequence for easier debugging.
file_seq_list = []
for key in sorted(seq2files.keys()):
file_seq_list.append(sorted(seq2files[key]))
logging.info('Loading {}: {}'.format(self.__class__.__name__,
self.DATA_PATH_FILE[phase]))
TemporalVoxelizationDataset.__init__(
self,
file_seq_list,
data_root=data_root,
input_transform=input_transform,
target_transform=target_transform,
ignore_label=config.ignore_label,
temporal_dilation=config.temporal_dilation,
temporal_numseq=config.temporal_numseq,
return_transformation=config.return_transformation,
augment_data=augment_data,
elastic_distortion=elastic_distortion,
config=config)
def __init__(self,
config,
input_transform=None,
target_transform=None,
augment_data=True,
cache=False,
phase=DatasetPhase.Train):
if isinstance(phase, str):
phase = str2datasetphase_type(phase)
data_root = config.sunrgbd_path
data_paths = read_txt(os.path.join(data_root, self.DATA_PATH_FILE[phase]))
logging.info('Loading {}: {}'.format(self.__class__.__name__, self.DATA_PATH_FILE[phase]))
super().__init__(
data_paths,
data_root=data_root,
input_transform=input_transform,
target_transform=target_transform,
ignore_label=config.ignore_label,
return_transformation=config.return_transformation,
augment_data=augment_data,
config=config)
import open3d as o3d
import lib.pc_utils as pc_utils
from config import get_config
from lib.utils import read_txt
SCANNET_RAW_PATH = '/cvgl2/u/jgwak/Datasets/scannet_raw'
SCANNET_ALIGNMENT_PATH = '/cvgl2/u/jgwak/Datasets/scannet_raw/scans/%s/%s.txt'
VOTENET_PRED_PATH = 'outputs/visualization/votenet_scannet'
OURS_PRED_PATH = 'outputs/visualization/ours_scannet'
SIS_PRED_PATH = 'outputs/visualization/3dsis_scannet'
NUM_BBOX_POINTS = 1000
config = get_config()
files = sorted(
read_txt(
'/scr/jgwak/Datasets/scannet_votenet_rgb/scannet_votenet_test.txt'))
for i, fn in enumerate(files):
filename = fn.split(os.sep)[-1][:-4]
if not os.path.isfile(os.path.join(SIS_PRED_PATH, f'{filename}.npz')):
continue
file_path = os.path.join(SCANNET_RAW_PATH, 'scans', filename,
f"{filename}_vh_clean.ply")
assert os.path.isfile(file_path)
mesh = o3d.io.read_triangle_mesh(file_path)
mesh.compute_vertex_normals()
scene_f = SCANNET_ALIGNMENT_PATH % (filename, filename)
alignment_txt = [
l for l in read_txt(scene_f) if l.startswith('axisAlignment = ')
][0]
rot = np.array([float(x)
for x in alignment_txt[16:].split()]).reshape(4, 4)
import os
import numpy as np
import lib.pc_utils as pc_utils
from lib.utils import read_txt
OURS_PRED_PATH = 'outputs/visualization/ours_stanford'
STANFORD_PATH = '/cvgl/group/Stanford3dDataset_v1.2/Area_5'
NUM_BBOX_POINTS = 1000
files = sorted(read_txt('/scr/jgwak/Datasets/stanford3d/test.txt'))
for i, fn in enumerate(files):
area_name = os.path.splitext(fn.split(os.sep)[-1])[0]
area_name = '_'.join(area_name.split('_')[1:])[:-2]
# area_name = 'office_34'
# i = [i for i, fn in enumerate(files) if area_name in fn][0]
if area_name.startswith('WC_') or area_name.startswith('hallway_'):
continue
ptc_fn = os.path.join(STANFORD_PATH, area_name, f'{area_name}.txt')
ptc = np.array([l.split() for l in read_txt(ptc_fn)]).astype(float)
pred_ours = np.load(os.path.join(OURS_PRED_PATH, 'out_%03d.npy.npz' % i))
gt = pc_utils.visualize_bboxes(pred_ours['gt'][:, :6], pred_ours['gt'][:, 6],
num_points=NUM_BBOX_POINTS)
pred_ours = pc_utils.visualize_bboxes(pred_ours['pred'][:, :6], pred_ours['pred'][:, 6],
num_points=NUM_BBOX_POINTS)
params = pc_utils.visualize_pcd(gt, ptc, save_image=f'viz_Area5_{area_name}_gt.png')
pc_utils.visualize_pcd(ptc, camera_params=params, save_image=f'viz_Area5_{area_name}_input.png')
pc_utils.visualize_pcd(pred_ours, ptc, camera_params=params,
save_image=f'viz_Area5_{area_name}_ours.png')
