print('[Data] outputs: %s' % ' x '.join(map(str, outputs.shape)))
print('[Data] inputs: %s' % ' x '.join(map(str, inputs.shape)))
print('[Data] space: %s' % ' x '.join(map(str, space.shape)))
n_observations = config['n_observations']
for i in range(min(25, outputs.shape[0])):
#n = random.randint(0, outputs[-1].shape[0])
n = i
print('[Data] visualizing %s %d/%d' % (set,
(n + 1), outputs.shape[0]))
for k in range(n_observations):
slices = []
for j in range(2, outputs[n][0].shape[0] - 2, 2):
slices.append(j)
outputs_inputs = outputs[n][0].copy()
outputs_inputs[inputs[n][k] == 1] = 2
space_inputs = space[n][k].copy()
space_inputs[inputs[n][k] == 1] = 2
mask = np.logical_and(inputs[n][k] == 1, space[n][k] == 1)
space_inputs[mask] = 3
visualization.plot_specific_slices([
outputs[n][0], inputs[n][k], space[n][k], outputs_inputs,
space_inputs
], slices, '%s_%d_%d.png' % (set, n, k), 0, 0, 3)
n = i
print('[Data] visualizing %s %d/%d' % (set,
(n + 1), filled.shape[0]))
for k in range(n_observations):
slices = []
for j in range(2, filled[n][0].shape[0] - 2, 2):
slices.append(j)
filled_inputs = filled[n][0].copy()
filled_inputs[inputs[n][k] == 1] = 2
space_inputs = space[n][k].copy()
space_inputs[inputs[n][k] == 1] = 2
mask = np.logical_and(inputs[n][k] == 1, space[n][k] == 1)
space_inputs[mask] = 3
if config['sdf']:
sdf_bin = sdfs[n][0].copy()
sdf_bin[sdf_bin >= 0] = 0
sdf_bin[sdf_bin < 0] = 1
visualization.plot_specific_slices([
filled[n][0], inputs[n][k], space[n][k], filled_inputs,
space_inputs, sdf_bin
], slices, '%s_%d_%d.png' % (set, n, k), 0, 0, 3)
else:
visualization.plot_specific_slices([
filled[n][0], inputs[n][k], space[n][k], filled_inputs,
space_inputs
], slices, '%s_%d_%d.png' % (set, n, k), 0, 0, 3)
#for level in range(config['levels'] + 1):
# volumes.append(outputs[level][n][0])
# volumes.append(filled[level][n][0])
#visualization.plot_slices(volumes, dataset + '_' + str(n) + '.png', 0, 0, np.max(labeled))
#plt.clf()
#plt.imshow(depths[n])
#plt.savefig(dataset + '_' + str(n) + '_depth.png')
slices = []
for j in range(2, filled[n][0].shape[0] - 2, 1):
slices.append(j)
filled_inputs = filled[n][0].copy()
filled_inputs[inputs[n][0] == 1] = 2
space_inputs = space[n][0].copy()
space_inputs[inputs[n][0] == 1] = 2
mask = np.logical_and(inputs[n][0] == 1, space[n][0] == 1)
space_inputs[mask] = 3
real_sdf_bin = real_sdf[n][0].copy()
real_sdf_bin[real_sdf_bin >= 0] = 0
real_sdf_bin[real_sdf_bin < 0] = 1
visualization.plot_specific_slices([
filled[n][0], inputs[n][0], space[n][0], filled_inputs,
space_inputs, real_sdf_bin
], slices, dataset + '_' + str(n) + '.png', 0, 0, 3)
#visualization.plot_specific_slices([real_sdf[n][0], input_sdf[n][0]], slices, dataset + '_' + str(n) + '_sdf.png', 0, min(np.min(real_sdf[n][0]), np.min(input_sdf[n][0])), max(np.max(real_sdf[n][0]), np.max(input_sdf[n][0])))
# print(' filtered (%d/%d %d/%d %f/%f' % (
# n_voxels, config['min_input_voxels'],
# n_space, config['min_space_voxels'],
# distance, config['max_bounding_box_distance']
# ))
#else:
# print(' not filtered (%d/%d %d/%d %f/%f' % (
# n_voxels, config['min_input_voxels'],
# n_space, config['min_space_voxels'],
# distance, config['max_bounding_box_distance']
# ))
slices = []
for j in range(2, inputs[n][0].shape[0] - 2, 1):
slices.append(j)
visualization.plot_specific_slices(
[inputs[n][0], full_space[n][0], part_space[n][0]], slices,
dataset + '_' + str(n) + '.png')
#visualization.plot_specific_slices([inputs_combined_gt[n][0], inputs_gt[n][0], full_space_gt[n][0], part_space_gt[n][0], inputs_gt[n][1], full_space_gt[n][1], part_space_gt[n][1], inputs_gt[n][2], full_space_gt[n][2], part_space_gt[n][2]], slices, dataset + '_' + str(n) + '_gt.png')
#visualization.plot_specific_slices([inputs_sdf_combined_gt[n][0], inputs_sdf[n][0]], slices, dataset + '_' + str(n) + '_sdf.png', 0, np.min(inputs_sdf[n]),np.max(inputs_sdf[n]))
#visualization.plot_volumes(space[n][0], inputs[n][0], '', 10, 1)
#visualization.plot_volume_mayavi(inputs[n][0])
#visualization.plot_volume_mayavi(full_space[n][0])
#visualization.plot_volume_mayavi(part_space[n][0])
#plt.imshow(depths[n, :, :, 0])
#plt.imshow(cylindral[n])
#plt.show()
print('[Data] read ' +
common.filename(config, 'input_file', '_f.h5', dataset))
inputs_sdf = utils.read_hdf5(
common.filename(config, 'input_sdf_file', '_f.h5', dataset))
print('[Data] read ' +
common.filename(config, 'input_sdf_file', '_f.h5', dataset))
inputs_tsdf = utils.read_hdf5(
common.filename(config, 'input_tsdf_file', '_f.h5', dataset))
print('[Data] read ' +
common.filename(config, 'input_tsdf_file', '_f.h5', dataset))
inputs_ltsdf = utils.read_hdf5(
common.filename(config, 'input_ltsdf_file', '_f.h5', dataset))
print('[Data] read ' +
common.filename(config, 'input_ltsdf_file', '_f.h5', dataset))
for i in range(min(25, inputs.shape[0])):
n = i #random.randint(0, inputs.shape[0] - 1)
print('[Data] visualizing %s %d/%d' % (dataset,
(n + 1), inputs.shape[0]))
slices = []
for j in range(2, inputs[n][0].shape[0] - 2, 1):
slices.append(j)
visualization.plot_specific_slices([
inputs[n][0] * np.max(inputs_sdf[n][0]), inputs_sdf[n][0],
inputs_tsdf[n][0], inputs_ltsdf[n][0]
], slices, dataset + '_' + str(n) + '.png', 0, 0,
np.max(inputs_sdf[n][0]))
input_sdfs = utils.read_hdf5(common.filename(config, 'input_sdf_file'))
print('[Data] read ' + common.filename(config, 'input_sdf_file'))
input_ltsdfs = utils.read_hdf5(common.filename(config, 'input_ltsdf_file'))
print('[Data] read ' + common.filename(config, 'input_ltsdf_file'))
output_sdfs = utils.read_hdf5(common.filename(config, 'sdf_file'))
print('[Data] reading ' + common.filename(config, 'sdf_file'))
output_ltsdfs = utils.read_hdf5(common.filename(config, 'ltsdf_file'))
print('[Data] reading ' + common.filename(config, 'ltsdf_file'))
print('[Data] input_sdfs: %s' % ' x '.join(map(str, input_sdfs.shape)))
print('[Data] input_ltsdfs: %s' % ' x '.join(map(str, input_ltsdfs.shape)))
print('[Data] output_sdfs: %s' % ' x '.join(map(str, output_sdfs.shape)))
print('[Data] output_ltsdfs: %s' % ' x '.join(map(str, output_ltsdfs.shape)))
for i in range(min(25, output_sdfs.shape[0])):
#n = random.randint(0, filled[-1].shape[0])
n = i
print('[Data] visualizing %s %d/%d' % (set, (n + 1), output_sdfs.shape[0]))
slices = []
for j in range(2, output_sdfs[n][0].shape[0] - 2, 2):
slices.append(j)
visualization.plot_specific_slices([
output_sdfs[n][0],
output_ltsdfs[n][0],
input_sdfs[n][0],
input_ltsdfs[n][0]
], slices, set + '_' + str(n) + '.png', 0, np.min(output_sdfs[n][0]), max(np.max(output_sdfs[n][0]), np.max(input_sdfs[n][0])))
print('[Data] Usage python 13_ply_observations.py config_folder')
exit(1)
config_folder = sys.argv[1] + '/'
assert os.path.exists(
config_folder), 'directory %s does not exist' % config_folder
config_file = config_folder + 'training_prior.json'
assert os.path.exists(config_file)
print('[Data] reading ' + config_file)
config = utils.read_json(config_file)
filled_file = common.filename(config, 'filled_file')
filled = utils.read_hdf5(filled_file)
print('[Data] read ' + filled_file)
slices = []
for j in range(2, filled[0][0].shape[0] - 2, 2):
slices.append(j)
N = filled.shape[0]
statistics = np.sum(np.squeeze(filled), axis=0) / float(N)
statistics = 1 - statistics
statistics = (0.75 * statistics + 0.25 * 1)**2
visualization.plot_specific_slices([statistics], slices)
statistics_file = common.filename(config, 'training_statistics_file')
utils.write_hdf5(statistics_file, statistics)
print('[Data] wrote ' + statistics_file)