def get_dataset_params(name, model_type='', train_type='', test_type='',
cam_type=''):
p = {'name': name, 'model_type': model_type,
'train_type': train_type, 'test_type': test_type, 'cam_type': cam_type}
# Path to the folder with datasets
common_base_path = '/home/meiqua/6DPose/public/datasets/'
# Path to the T-LESS Toolkit (https://github.com/thodan/t-less_toolkit)
tless_tk_path = '../t_less_toolkit/'
if name == 'hinterstoisser':
p['obj_count'] = 15
p['scene_count'] = 15
p['train_im_size'] = (640, 480)
p['test_im_size'] = (640, 480)
p['base_path'] = pjoin(common_base_path, 'hinterstoisser')
p['im_id_pad'] = 4
p['model_texture_mpath'] = None
p['cam_params_path'] = pjoin(p['base_path'], 'camera.yml')
# p['test_obj_depth_range'] = (600.90, 1102.35) # [mm] - with original GT
p['test_obj_depth_range'] = (346.31, 1499.84) # [mm] - with extended GT
# (there are only 3 occurrences under 400 mm)
p['test_obj_azimuth_range'] = (0, 2 * math.pi)
p['test_obj_elev_range'] = (0, 0.5 * math.pi)
elif name == 'tless':
p['obj_count'] = 30
p['scene_count'] = 20
p['base_path'] = pjoin(common_base_path, 't-less', 't-less_v2')
p['im_id_pad'] = 4
p['model_texture_mpath'] = None
if p['model_type'] == '': p['model_type'] = 'cad'
if p['train_type'] == '': p['train_type'] = 'primesense'
if p['test_type'] == '': p['test_type'] = 'primesense'
if p['cam_type'] == '': p['cam_type'] = 'primesense'
p['cam_params_path'] = pjoin(tless_tk_path, 'cam',
'camera_' + p['cam_type'] + '.yml')
if p['test_type'] in ['primesense', 'kinect']:
p['test_im_size'] = (720, 540)
elif p['test_type'] == 'canon':
p['test_im_size'] = (2560, 1920)
if p['train_type'] in ['primesense', 'kinect']:
p['train_im_size'] = (400, 400)
elif p['train_type'] == 'canon':
p['train_im_size'] = (1900, 1900)
elif p['train_type'] == 'render_reconst':
p['train_im_size'] = (1280, 1024)
if p['test_type'] == 'primesense':
p['test_obj_depth_range'] = (649.89, 940.04) # [mm]
p['test_obj_azimuth_range'] = (0, 2 * math.pi)
p['test_obj_elev_range'] = (-0.5 * math.pi, 0.5 * math.pi)
if p['test_type'] == 'kinect':
# Not calculated yet
p['test_depth_range'] = None
p['test_obj_azimuth_range'] = None
p['test_obj_elev_range'] = None
elif p['test_type'] == 'canon':
# Not calculated yet
p['test_depth_range'] = None
p['test_obj_azimuth_range'] = None
p['test_obj_elev_range'] = None
elif name == 'tudlight':
p['obj_count'] = 3
p['scene_count'] = 3
p['train_im_size'] = (640, 480)
p['test_im_size'] = (640, 480)
p['base_path'] = pjoin(common_base_path, 'tudlight')
p['im_id_pad'] = 5 # 5
p['model_texture_mpath'] = None
p['cam_params_path'] = pjoin(p['base_path'], 'camera.yml')
p['test_obj_depth_range'] = (851.29, 2016.14) # [mm]
p['test_obj_azimuth_range'] = (0, 2 * math.pi)
p['test_obj_elev_range'] = (-0.4363, 0.5 * math.pi) # (-25, 90) [deg]
elif name == 'toyotalight':
p['obj_count'] = 21
p['scene_count'] = 21
p['train_im_size'] = (640, 480)
p['test_im_size'] = (640, 480)
p['base_path'] = pjoin(common_base_path, 'toyotalight')
p['im_id_pad'] = 4 # 5
p['model_texture_mpath'] = None
p['cam_params_path'] = pjoin(p['base_path'], 'camera.yml')
# p['test_obj_depth_range'] = None # [mm]
# p['test_obj_azimuth_range'] = None
# p['test_obj_elev_range'] = None # (-25, 90) [deg]
elif name == 'rutgers':
p['obj_count'] = 14
p['scene_count'] = 14
p['train_im_size'] = (640, 480)
p['test_im_size'] = (640, 480)
p['base_path'] = pjoin(common_base_path, 'rutgers')
p['im_id_pad'] = 4
p['model_texture_mpath'] = pjoin(p['base_path'], 'models', 'obj_{:02d}.png')
p['cam_params_path'] = pjoin(p['base_path'], 'camera.yml')
p['test_obj_depth_range'] = (594.41, 739.12) # [mm]
p['test_obj_azimuth_range'] = (0, 2 * math.pi)
p['test_obj_elev_range'] = (-0.5 * math.pi, 0.5 * math.pi)
elif name == 'tejani':
p['obj_count'] = 6
p['scene_count'] = 6
p['train_im_size'] = (640, 480)
p['test_im_size'] = (640, 480)
p['base_path'] = pjoin(common_base_path, 'tejani')
p['im_id_pad'] = 4
p['model_texture_mpath'] = None
p['cam_params_path'] = pjoin(p['base_path'], 'camera.yml')
p['test_obj_depth_range'] = (509.12 - 1120.41) # [mm]
p['test_obj_azimuth_range'] = (0, 2 * math.pi)
p['test_obj_elev_range'] = (0, 0.5 * math.pi)
elif name == 'doumanoglou':
p['obj_count'] = 2
p['scene_count'] = 3
p['train_im_size'] = (640, 480)
p['test_im_size'] = (640, 480)
p['base_path'] = pjoin(common_base_path, 'doumanoglou')
p['im_id_pad'] = 4
p['model_texture_mpath'] = None
p['cam_params_path'] = pjoin(p['base_path'], 'camera.yml')
p['test_obj_depth_range'] = (454.56 - 1076.29) # [mm]
p['test_obj_azimuth_range'] = (0, 2 * math.pi)
p['test_obj_elev_range'] = (-1.0297, 0.5 * math.pi) # (-59, 90) [deg]
else:
print('Error: Unknown SIXD dataset.')
exit(-1)
models_dir = 'models' if p['model_type'] == '' else 'models_' + p['model_type']
train_dir = 'train' if p['train_type'] == '' else 'train_' + p['train_type']
test_dir = 'test' if p['test_type'] == '' else 'test_' + p['test_type']
# Image ID format
im_id_f = '{:' + str(p['im_id_pad']).zfill(2) + 'd}'
# Paths and path masks
p['model_mpath'] = pjoin(p['base_path'], models_dir, 'obj_{:02d}.ply')
p['models_info_path'] = pjoin(p['base_path'], models_dir, 'models_info.yml')
p['obj_info_mpath'] = pjoin(p['base_path'], train_dir, '{:02d}', 'info.yml')
p['obj_gt_mpath'] = pjoin(p['base_path'], train_dir, '{:02d}', 'gt.yml')
p['obj_gt_stats_mpath'] = pjoin(p['base_path'], train_dir + '_gt_stats', '{:02d}_delta={}.yml')
p['train_rgb_mpath'] = pjoin(p['base_path'], train_dir, '{:02d}', 'rgb', im_id_f + '.png')
p['train_depth_mpath'] = pjoin(p['base_path'], train_dir, '{:02d}', 'depth', im_id_f + '.png')
p['train_mask_mpath'] = pjoin(p['base_path'], train_dir, '{:02d}', 'mask', im_id_f + '_{:02d}.png')
p['train_mask_visib_mpath'] = pjoin(p['base_path'], train_dir, '{:02d}', 'mask_visib', im_id_f + '_{:02d}.png')
p['scene_info_mpath'] = pjoin(p['base_path'], test_dir, '{:02d}', 'info.yml')
p['scene_gt_mpath'] = pjoin(p['base_path'], test_dir, '{:02d}', 'gt.yml')
p['scene_gt_stats_mpath'] = pjoin(p['base_path'], test_dir + '_gt_stats', '{:02d}_delta={}.yml')
p['test_rgb_mpath'] = pjoin(p['base_path'], test_dir, '{:02d}', 'rgb', im_id_f + '.png')
p['test_depth_mpath'] = pjoin(p['base_path'], test_dir, '{:02d}', 'depth', im_id_f + '.png')
p['test_mask_mpath'] = pjoin(p['base_path'], test_dir, '{:02d}', 'mask', im_id_f + '_{:02d}.png')
p['test_mask_visib_mpath'] = pjoin(p['base_path'], test_dir, '{:02d}', 'mask_visib', im_id_f + '_{:02d}.png')
p['test_set_fpath'] = pjoin(p['base_path'], 'test_set_v1.yml')
p['cam'] = inout.load_cam_params(p['cam_params_path'])
return p
p = dict()
p['name'] = 'customDataset'
p['obj_count'] = 2
p['scene_count'] = 2
p['train_im_size'] = (640, 480)
p['test_im_size'] = (640, 480)
p['base_path'] = join('/home/meiqua/6DPose/public/datasets/', p['name'])
misc.ensure_dir(p['base_path'])
p['model_mpath'] = join(p['base_path'], 'models', 'obj_{:02d}.ply')
misc.ensure_dir(os.path.dirname(p['model_mpath']))
p['model_texture_mpath'] = join(p['base_path'], 'models', 'obj_{:02d}.png')
p['cam_params_path'] = join(p['base_path'], 'camera.yml')
p['cam'] = inout.load_cam_params(p['cam_params_path'])
# Minimum required number of views on the whole view sphere. The final number of
# views depends on the sampling method.
min_n_views = 100
clip_near = 10 # [mm]
clip_far = 10000 # [mm]
ambient_weight = 0.8 # Weight of ambient light [0, 1]
shading = 'phong' # 'flat', 'phong'
# Super-sampling anti-aliasing (SSAA)
# https://github.com/vispy/vispy/wiki/Tech.-Antialiasing
# The RGB image is rendered at ssaa_fact times higher resolution and then
# down-sampled to the required resolution.
ssaa_fact = 4
def get_dataset_params(name, model_type='', train_type='', test_type='',
cam_type=''):
p = {'name': name, 'model_type': model_type,
'train_type': train_type, 'test_type': test_type, 'cam_type': cam_type}
# Path to the folder with datasets
common_base_path = pjoin(top_level_path, 'public/datasets/')
# Path to the T-LESS Toolkit (https://github.com/thodan/t-less_toolkit)
tless_tk_path = '../t_less_toolkit/'
if name == 'hinterstoisser':
p['obj_count'] = 15
p['scene_count'] = 15
p['train_im_size'] = (640, 480)
p['test_im_size'] = (640, 480)
p['base_path'] = pjoin(common_base_path, 'hinterstoisser')
p['im_id_pad'] = 4
p['model_texture_mpath'] = None
p['cam_params_path'] = pjoin(p['base_path'], 'camera.yml')
# p['test_obj_depth_range'] = (600.90, 1102.35) # [mm] - with original GT
p['test_obj_depth_range'] = (346.31, 1499.84) # [mm] - with extended GT
# (there are only 3 occurrences under 400 mm)
p['test_obj_azimuth_range'] = (0, 2 * math.pi)
p['test_obj_elev_range'] = (0, 0.5 * math.pi)
elif name == 'tless':
p['obj_count'] = 30
p['scene_count'] = 20
p['base_path'] = pjoin(common_base_path, 't-less', 't-less_v2')
p['im_id_pad'] = 4
p['model_texture_mpath'] = None
if p['model_type'] == '': p['model_type'] = 'cad'
if p['train_type'] == '': p['train_type'] = 'primesense'
if p['test_type'] == '': p['test_type'] = 'primesense'
if p['cam_type'] == '': p['cam_type'] = 'primesense'
p['cam_params_path'] = pjoin(tless_tk_path, 'cam',
'camera_' + p['cam_type'] + '.yml')
if p['test_type'] in ['primesense', 'kinect']:
p['test_im_size'] = (720, 540)
elif p['test_type'] == 'canon':
p['test_im_size'] = (2560, 1920)
if p['train_type'] in ['primesense', 'kinect']:
p['train_im_size'] = (400, 400)
elif p['train_type'] == 'canon':
p['train_im_size'] = (1900, 1900)
elif p['train_type'] == 'render_reconst':
p['train_im_size'] = (1280, 1024)
if p['test_type'] == 'primesense':
p['test_obj_depth_range'] = (649.89, 940.04) # [mm]
p['test_obj_azimuth_range'] = (0, 2 * math.pi)
p['test_obj_elev_range'] = (-0.5 * math.pi, 0.5 * math.pi)
if p['test_type'] == 'kinect':
# Not calculated yet
p['test_depth_range'] = None
p['test_obj_azimuth_range'] = None
p['test_obj_elev_range'] = None
elif p['test_type'] == 'canon':
# Not calculated yet
p['test_depth_range'] = None
p['test_obj_azimuth_range'] = None
p['test_obj_elev_range'] = None
elif name == 'tudlight':
p['obj_count'] = 3
p['scene_count'] = 3
p['train_im_size'] = (640, 480)
p['test_im_size'] = (640, 480)
p['base_path'] = pjoin(common_base_path, 'tudlight')
p['im_id_pad'] = 5 # 5
p['model_texture_mpath'] = None
p['cam_params_path'] = pjoin(p['base_path'], 'camera.yml')
p['test_obj_depth_range'] = (851.29, 2016.14) # [mm]
p['test_obj_azimuth_range'] = (0, 2 * math.pi)
p['test_obj_elev_range'] = (-0.4363, 0.5 * math.pi) # (-25, 90) [deg]
elif name == 'toyotalight':
p['obj_count'] = 21
p['scene_count'] = 21
p['train_im_size'] = (640, 480)
p['test_im_size'] = (640, 480)
p['base_path'] = pjoin(common_base_path, 'toyotalight')
p['im_id_pad'] = 4 # 5
p['model_texture_mpath'] = None
p['cam_params_path'] = pjoin(p['base_path'], 'camera.yml')
# p['test_obj_depth_range'] = None # [mm]
# p['test_obj_azimuth_range'] = None
# p['test_obj_elev_range'] = None # (-25, 90) [deg]
elif name == 'rutgers':
p['obj_count'] = 14
p['scene_count'] = 14
p['train_im_size'] = (640, 480)
p['test_im_size'] = (640, 480)
p['base_path'] = pjoin(common_base_path, 'rutgers')
p['im_id_pad'] = 4
p['model_texture_mpath'] = pjoin(p['base_path'], 'models', 'obj_{:02d}.png')
p['cam_params_path'] = pjoin(p['base_path'], 'camera.yml')
p['test_obj_depth_range'] = (594.41, 739.12) # [mm]
p['test_obj_azimuth_range'] = (0, 2 * math.pi)
p['test_obj_elev_range'] = (-0.5 * math.pi, 0.5 * math.pi)
elif name == 'tejani':
p['obj_count'] = 6
p['scene_count'] = 6
p['train_im_size'] = (640, 480)
p['test_im_size'] = (640, 480)
p['base_path'] = pjoin(common_base_path, 'tejani')
p['im_id_pad'] = 4
p['model_texture_mpath'] = None
p['cam_params_path'] = pjoin(p['base_path'], 'camera.yml')
p['test_obj_depth_range'] = (509.12 - 1120.41) # [mm]
p['test_obj_azimuth_range'] = (0, 2 * math.pi)
p['test_obj_elev_range'] = (0, 0.5 * math.pi)
elif name == 'doumanoglou':
p['obj_count'] = 2
p['scene_count'] = 3
p['train_im_size'] = (640, 480)
p['test_im_size'] = (640, 480)
p['base_path'] = pjoin(common_base_path, 'doumanoglou')
p['im_id_pad'] = 4
p['model_texture_mpath'] = None
p['cam_params_path'] = pjoin(p['base_path'], 'camera.yml')
p['test_obj_depth_range'] = (454.56 - 1076.29) # [mm]
p['test_obj_azimuth_range'] = (0, 2 * math.pi)
p['test_obj_elev_range'] = (-1.0297, 0.5 * math.pi) # (-59, 90) [deg]
else:
print('Error: Unknown SIXD dataset.')
exit(-1)
models_dir = 'models' if p['model_type'] == '' else 'models_' + p['model_type']
train_dir = 'train' if p['train_type'] == '' else 'train_' + p['train_type']
test_dir = 'test' if p['test_type'] == '' else 'test_' + p['test_type']
# Image ID format
im_id_f = '{:' + str(p['im_id_pad']).zfill(2) + 'd}'
# Paths and path masks
p['model_mpath'] = pjoin(p['base_path'], models_dir, 'obj_{:02d}.ply')
p['models_info_path'] = pjoin(p['base_path'], models_dir, 'models_info.yml')
p['obj_info_mpath'] = pjoin(p['base_path'], train_dir, '{:02d}', 'info.yml')
p['obj_gt_mpath'] = pjoin(p['base_path'], train_dir, '{:02d}', 'gt.yml')
p['obj_gt_stats_mpath'] = pjoin(p['base_path'], train_dir + '_gt_stats', '{:02d}_delta={}.yml')
p['train_rgb_mpath'] = pjoin(p['base_path'], train_dir, '{:02d}', 'rgb', im_id_f + '.png')
p['train_depth_mpath'] = pjoin(p['base_path'], train_dir, '{:02d}', 'depth', im_id_f + '.png')
p['train_mask_mpath'] = pjoin(p['base_path'], train_dir, '{:02d}', 'mask', im_id_f + '_{:02d}.png')
p['train_mask_visib_mpath'] = pjoin(p['base_path'], train_dir, '{:02d}', 'mask_visib', im_id_f + '_{:02d}.png')
p['scene_info_mpath'] = pjoin(p['base_path'], test_dir, '{:02d}', 'info.yml')
p['scene_gt_mpath'] = pjoin(p['base_path'], test_dir, '{:02d}', 'gt.yml')
p['scene_gt_stats_mpath'] = pjoin(p['base_path'], test_dir + '_gt_stats', '{:02d}_delta={}.yml')
p['test_rgb_mpath'] = pjoin(p['base_path'], test_dir, '{:02d}', 'rgb', im_id_f + '.png')
p['test_depth_mpath'] = pjoin(p['base_path'], test_dir, '{:02d}', 'depth', im_id_f + '.png')
p['test_mask_mpath'] = pjoin(p['base_path'], test_dir, '{:02d}', 'mask', im_id_f + '_{:02d}.png')
p['test_mask_visib_mpath'] = pjoin(p['base_path'], test_dir, '{:02d}', 'mask_visib', im_id_f + '_{:02d}.png')
p['test_set_fpath'] = pjoin(p['base_path'], 'test_set_v1.yml')
p['cam'] = inout.load_cam_params(p['cam_params_path'])
return p
def get_dataset_params(name,
model_type='',
train_type='',
test_type='',
cam_type=''):
p = {
'name': name,
'model_type': model_type,
'train_type': train_type,
'test_type': test_type,
'cam_type': cam_type
}
if name == 'hinterstoisser':
p['obj_count'] = 3
p['scene_count'] = 3
p['test_im_size'] = (640, 480)
p['base_path'] = '/local/datasets/tlod/hinterstoisser/'
p['im_id_pad'] = 4
p['model_texture_mpath'] = None
p['cam_params_path'] = p['base_path'] + 'camera.yml'
elif name == 'rutgers':
p['obj_count'] = 14
p['scene_count'] = 14
p['test_im_size'] = (640, 480)
p['base_path'] = '/local/datasets/tlod/rutgers/'
p['im_id_pad'] = 4
p['model_texture_mpath'] = p['base_path'] + 'models/obj_{:02d}.png'
p['cam_params_path'] = p['base_path'] + 'camera.yml'
elif name == 'tejani':
p['obj_count'] = 6
p['scene_count'] = 6
p['test_im_size'] = (640, 480)
p['base_path'] = '/local/datasets/tlod/imperial/tejani/'
p['im_id_pad'] = 4
p['model_texture_mpath'] = None
p['cam_params_path'] = p['base_path'] + 'camera.yml'
elif name == 'tless':
p['obj_count'] = 30
p['scene_count'] = 20
p['base_path'] = '/local/datasets/tlod/t-less/t-less_v2/'
p['im_id_pad'] = 4
p['model_texture_mpath'] = None
if p['model_type'] == '': p['model_type'] = 'cad'
if p['train_type'] == '': p['train_type'] = 'primesense'
if p['test_type'] == '': p['test_type'] = 'primesense'
if p['cam_type'] == '': p['cam_type'] = 'primesense'
p['cam_params_path'] = '/home/tom/th_data/cmp/projects/t-less/' \
't-less_toolkit/cam/camera_' + p['cam_type'] + '.yml'
if p['test_type'] in ['primesense', 'kinect']:
p['test_im_size'] = (720, 540)
elif p['test_type'] == 'canon':
p['test_im_size'] = (2560, 1920)
elif name == 'tudlight':
p['obj_count'] = 3
p['scene_count'] = 3
p['test_im_size'] = (640, 480)
p['base_path'] = '/local/datasets/tlod/dresden/tudlight/'
p['im_id_pad'] = 5
p['model_texture_mpath'] = None
p['cam_params_path'] = p['base_path'] + 'camera.yml'
models_dir = 'models' if p[
'model_type'] == '' else 'models_' + p['model_type']
train_dir = 'train' if p['train_type'] == '' else 'train_' + p['train_type']
test_dir = 'test' if p['test_type'] == '' else 'test_' + p['test_type']
# Image ID format
im_id_f = '{:' + str(p['im_id_pad']).zfill(2) + 'd}'
# Paths and path masks
p['model_mpath'] = p['base_path'] + models_dir + '/obj_{:02d}.ply'
p['models_info_path'] = p['base_path'] + models_dir + '/models_info.yml'
p['obj_info_mpath'] = p['base_path'] + train_dir + '/{:02d}/info.yml'
p['obj_gt_mpath'] = p['base_path'] + train_dir + '/{:02d}/gt.yml'
p['train_rgb_mpath'] = p[
'base_path'] + train_dir + '/{:02d}/rgb/' + im_id_f + '.png'
p['train_depth_mpath'] = p[
'base_path'] + train_dir + '/{:02d}/depth/' + im_id_f + '.png'
p['scene_info_mpath'] = p['base_path'] + test_dir + '/{:02d}/info.yml'
p['scene_gt_mpath'] = p['base_path'] + test_dir + '/{:02d}/gt.yml'
p['test_rgb_mpath'] = p[
'base_path'] + test_dir + '/{:02d}/rgb/' + im_id_f + '.png'
p['test_depth_mpath'] = p[
'base_path'] + test_dir + '/{:02d}/depth/' + im_id_f + '.png'
p['cam'] = inout.load_cam_params(p['cam_params_path'])
return p
def get_dataset_params(name,
model_type='',
train_type='',
test_type='',
cam_type=''):
p = {
'name': name,
'model_type': model_type,
'train_type': train_type,
'test_type': test_type,
'cam_type': cam_type
}
# Folder with datasets
common_base_path = '/local/datasets/tlod/'
# common_base_path = '/datagrid/personal/hodanto2/datasets/sixd/'
# Path to the T-LESS Toolkit (https://github.com/thodan/t-less_toolkit)
tless_tk_path = '/home/tom/th_data/cmp/projects/t-less/t-less_toolkit/'
# tless_tk_path = '/home.dokt/hodanto2/projects/t-less/t-less_toolkit/'
if name == 'hinterstoisser':
p['obj_count'] = 15
p['scene_count'] = 15
p['train_im_size'] = (640, 480)
p['test_im_size'] = (640, 480)
p['base_path'] = common_base_path + 'hinterstoisser/'
p['im_id_pad'] = 4
p['model_texture_mpath'] = None
p['cam_params_path'] = p['base_path'] + 'camera.yml'
elif name == 'tless':
p['obj_count'] = 30
p['scene_count'] = 20
p['base_path'] = common_base_path + 't-less/t-less_v2/'
p['im_id_pad'] = 4
p['model_texture_mpath'] = None
if p['model_type'] == '': p['model_type'] = 'cad'
if p['train_type'] == '': p['train_type'] = 'primesense'
if p['test_type'] == '': p['test_type'] = 'primesense'
if p['cam_type'] == '': p['cam_type'] = 'primesense'
p['cam_params_path'] = tless_tk_path + 'cam/camera_' + p[
'cam_type'] + '.yml'
if p['test_type'] in ['primesense', 'kinect']:
p['train_im_size'] = (400, 400)
p['test_im_size'] = (720, 540)
elif p['test_type'] == 'canon':
p['train_im_size'] = (1900, 1900)
p['test_im_size'] = (2560, 1920)
elif name == 'tudlight':
p['obj_count'] = 3
p['scene_count'] = 3
p['train_im_size'] = (640, 480)
p['test_im_size'] = (640, 480)
p['base_path'] = common_base_path + 'tudlight/'
p['im_id_pad'] = 5
p['model_texture_mpath'] = None
p['cam_params_path'] = p['base_path'] + 'camera.yml'
elif name == 'rutgers':
p['obj_count'] = 14
p['scene_count'] = 14
p['train_im_size'] = (640, 480)
p['test_im_size'] = (640, 480)
p['base_path'] = common_base_path + 'rutgers/'
p['im_id_pad'] = 4
p['model_texture_mpath'] = p['base_path'] + 'models/obj_{:02d}.png'
p['cam_params_path'] = p['base_path'] + 'camera.yml'
elif name == 'tejani':
p['obj_count'] = 6
p['scene_count'] = 6
p['train_im_size'] = (640, 480)
p['test_im_size'] = (640, 480)
p['base_path'] = common_base_path + 'tejani/'
p['im_id_pad'] = 4
p['model_texture_mpath'] = None
p['cam_params_path'] = p['base_path'] + 'camera.yml'
elif name == 'doumanoglou':
p['obj_count'] = 2
p['scene_count'] = 3
p['train_im_size'] = (640, 480)
p['test_im_size'] = (640, 480)
p['base_path'] = common_base_path + 'doumanoglou/'
p['im_id_pad'] = 4
p['model_texture_mpath'] = None
p['cam_params_path'] = p['base_path'] + 'camera.yml'
else:
print('Error: Unknown SIXD dataset.')
exit(-1)
models_dir = 'models' if p[
'model_type'] == '' else 'models_' + p['model_type']
train_dir = 'train' if p['train_type'] == '' else 'train_' + p['train_type']
test_dir = 'test' if p['test_type'] == '' else 'test_' + p['test_type']
# Image ID format
im_id_f = '{:' + str(p['im_id_pad']).zfill(2) + 'd}'
# Paths and path masks
p['model_mpath'] = p['base_path'] + models_dir + '/obj_{:02d}.ply'
p['models_info_path'] = p['base_path'] + models_dir + '/models_info.yml'
p['obj_info_mpath'] = p['base_path'] + train_dir + '/{:02d}/info.yml'
p['obj_gt_mpath'] = p['base_path'] + train_dir + '/{:02d}/gt.yml'
p['train_rgb_mpath'] = p[
'base_path'] + train_dir + '/{:02d}/rgb/' + im_id_f + '.png'
p['train_depth_mpath'] = p[
'base_path'] + train_dir + '/{:02d}/depth/' + im_id_f + '.png'
p['scene_info_mpath'] = p['base_path'] + test_dir + '/{:02d}/info.yml'
p['scene_gt_mpath'] = p['base_path'] + test_dir + '/{:02d}/gt.yml'
p['scene_gt_stats_mpath'] = p[
'base_path'] + test_dir + '_gt_stats/{:02d}_delta={}.yml'
p['test_rgb_mpath'] = p[
'base_path'] + test_dir + '/{:02d}/rgb/' + im_id_f + '.png'
p['test_depth_mpath'] = p[
'base_path'] + test_dir + '/{:02d}/depth/' + im_id_f + '.png'
p['cam'] = inout.load_cam_params(p['cam_params_path'])
return p
p = dict()
p['name'] = 'customDataset'
p['obj_count'] = 2
p['scene_count'] = 2
p['train_im_size'] = (640, 480)
p['test_im_size'] = (640, 480)
p['base_path'] = join('/home/meiqua/6DPose/public/datasets/', p['name'])
misc.ensure_dir(p['base_path'])
p['model_mpath'] = join(p['base_path'], 'models', 'obj_{:02d}.ply')
misc.ensure_dir(os.path.dirname(p['model_mpath']))
p['model_texture_mpath'] = join(p['base_path'], 'models', 'obj_{:02d}.png')
p['cam_params_path'] = join(p['base_path'], 'camera.yml')
p['cam'] = inout.load_cam_params(p['cam_params_path'])
# Minimum required number of views on the whole view sphere. The final number of
# views depends on the sampling method.
min_n_views = 100
clip_near = 10 # [mm]
clip_far = 10000 # [mm]
ambient_weight = 0.8 # Weight of ambient light [0, 1]
shading = 'phong' # 'flat', 'phong'
# Super-sampling anti-aliasing (SSAA)
# https://github.com/vispy/vispy/wiki/Tech.-Antialiasing
# The RGB image is rendered at ssaa_fact times higher resolution and then
# down-sampled to the required resolution.
ssaa_fact = 4
