def test_without_placeholder(self):
test_labels = ArrayKey("TEST_LABELS")
test_points = GraphKey("TEST_POINTS")
pipeline = (
PointTestSource3D() + RandomLocation(ensure_nonempty=test_points) +
ElasticAugment([10, 10, 10], [0.1, 0.1, 0.1], [0, 2.0 * math.pi]) +
Snapshot(
{test_labels: "volumes/labels"},
output_dir=self.path_to(),
output_filename="elastic_augment_test{id}-{iteration}.hdf",
))
with build(pipeline):
for i in range(2):
request_size = Coordinate((40, 40, 40))
request_a = BatchRequest(random_seed=i)
request_a.add(test_points, request_size)
request_b = BatchRequest(random_seed=i)
request_b.add(test_points, request_size)
request_b.add(test_labels, request_size)
# No array to provide a voxel size to ElasticAugment
with pytest.raises(PipelineRequestError):
pipeline.request_batch(request_a)
batch_b = pipeline.request_batch(request_b)
self.assertIn(test_labels, batch_b)
def test_placeholder(self):
test_labels = ArrayKey("TEST_LABELS")
test_points = GraphKey("TEST_POINTS")
pipeline = (
PointTestSource3D() + RandomLocation(ensure_nonempty=test_points) +
ElasticAugment([10, 10, 10], [0.1, 0.1, 0.1], [0, 2.0 * math.pi]) +
Snapshot(
{test_labels: "volumes/labels"},
output_dir=self.path_to(),
output_filename="elastic_augment_test{id}-{iteration}.hdf",
))
with build(pipeline):
for i in range(2):
request_size = Coordinate((40, 40, 40))
request_a = BatchRequest(random_seed=i)
request_a.add(test_points, request_size)
request_a.add(test_labels, request_size, placeholder=True)
request_b = BatchRequest(random_seed=i)
request_b.add(test_points, request_size)
request_b.add(test_labels, request_size)
batch_a = pipeline.request_batch(request_a)
batch_b = pipeline.request_batch(request_b)
points_a = batch_a[test_points].nodes
points_b = batch_b[test_points].nodes
for a, b in zip(points_a, points_b):
assert all(np.isclose(a.location, b.location))
def test_3d_basics(self):
test_labels = ArrayKey("TEST_LABELS")
test_points = GraphKey("TEST_POINTS")
test_raster = ArrayKey("TEST_RASTER")
pipeline = (
PointTestSource3D() + ElasticAugment(
[10, 10, 10],
[0.1, 0.1, 0.1],
# [0, 0, 0], # no jitter
[0, 2.0 * math.pi],
) + RasterizeGraph(
test_points,
test_raster,
settings=RasterizationSettings(radius=2, mode="peak"),
) + Snapshot(
{
test_labels: "volumes/labels",
test_raster: "volumes/raster"
},
dataset_dtypes={test_raster: np.float32},
output_dir=self.path_to(),
output_filename="elastic_augment_test{id}-{iteration}.hdf",
))
for _ in range(5):
with build(pipeline):
request_roi = Roi((-20, -20, -20), (40, 40, 40))
request = BatchRequest()
request[test_labels] = ArraySpec(roi=request_roi)
request[test_points] = GraphSpec(roi=request_roi)
request[test_raster] = ArraySpec(roi=request_roi)
batch = pipeline.request_batch(request)
labels = batch[test_labels]
points = batch[test_points]
# the point at (0, 0, 0) should not have moved
self.assertTrue(points.contains(0))
labels_data_roi = (
labels.spec.roi -
labels.spec.roi.get_begin()) / labels.spec.voxel_size
# points should have moved together with the voxels
for point in points.nodes:
loc = point.location - labels.spec.roi.get_begin()
loc = loc / labels.spec.voxel_size
loc = Coordinate(int(round(x)) for x in loc)
if labels_data_roi.contains(loc):
self.assertEqual(labels.data[loc], point.id)
def train():
gunpowder.set_verbose(False)
affinity_neighborhood = malis.mknhood3d()
solver_parameters = gunpowder.caffe.SolverParameters()
solver_parameters.train_net = 'net.prototxt'
solver_parameters.base_lr = 1e-4
solver_parameters.momentum = 0.95
solver_parameters.momentum2 = 0.999
solver_parameters.delta = 1e-8
solver_parameters.weight_decay = 0.000005
solver_parameters.lr_policy = 'inv'
solver_parameters.gamma = 0.0001
solver_parameters.power = 0.75
solver_parameters.snapshot = 10000
solver_parameters.snapshot_prefix = 'net'
solver_parameters.type = 'Adam'
solver_parameters.resume_from = None
solver_parameters.train_state.add_stage('euclid')
request = BatchRequest()
request.add_volume_request(VolumeTypes.RAW, constants.input_shape)
request.add_volume_request(VolumeTypes.GT_LABELS, constants.output_shape)
request.add_volume_request(VolumeTypes.GT_MASK, constants.output_shape)
request.add_volume_request(VolumeTypes.GT_AFFINITIES, constants.output_shape)
request.add_volume_request(VolumeTypes.LOSS_SCALE, constants.output_shape)
data_providers = list()
fibsem_dir = "/groups/turaga/turagalab/data/FlyEM/fibsem_medulla_7col"
for volume_name in ("tstvol-520-1-h5",):
h5_filepath = "./{}.h5".format(volume_name)
path_to_labels = os.path.join(fibsem_dir, volume_name, "groundtruth_seg.h5")
with h5py.File(path_to_labels, "r") as f_labels:
mask_shape = f_labels["main"].shape
with h5py.File(h5_filepath, "w") as h5:
h5['volumes/raw'] = h5py.ExternalLink(os.path.join(fibsem_dir, volume_name, "im_uint8.h5"), "main")
h5['volumes/labels/neuron_ids'] = h5py.ExternalLink(path_to_labels, "main")
h5.create_dataset(
name="volumes/labels/mask",
dtype="uint8",
shape=mask_shape,
fillvalue=1,
)
data_providers.append(
gunpowder.Hdf5Source(
h5_filepath,
datasets={
VolumeTypes.RAW: 'volumes/raw',
VolumeTypes.GT_LABELS: 'volumes/labels/neuron_ids',
VolumeTypes.GT_MASK: 'volumes/labels/mask',
},
resolution=(8, 8, 8),
)
)
dvid_source = DvidSource(
hostname='slowpoke3',
port=32788,
uuid='341',
raw_array_name='grayscale',
gt_array_name='groundtruth',
gt_mask_roi_name="seven_column_eroded7_z_lt_5024",
resolution=(8, 8, 8),
)
data_providers.extend([dvid_source])
data_providers = tuple(
provider +
RandomLocation() +
Reject(min_masked=0.5) +
Normalize()
for provider in data_providers
)
# create a batch provider by concatenation of filters
batch_provider = (
data_providers +
RandomProvider() +
ElasticAugment([20, 20, 20], [0, 0, 0], [0, math.pi / 2.0]) +
SimpleAugment(transpose_only_xy=False) +
GrowBoundary(steps=2, only_xy=False) +
AddGtAffinities(affinity_neighborhood) +
BalanceAffinityLabels() +
SplitAndRenumberSegmentationLabels() +
IntensityAugment(0.9, 1.1, -0.1, 0.1, z_section_wise=False) +
PreCache(
request,
cache_size=11,
num_workers=10) +
Train(solver_parameters, use_gpu=0) +
Typecast(volume_dtypes={
VolumeTypes.GT_LABELS: np.dtype("uint32"),
VolumeTypes.GT_MASK: np.dtype("uint8"),
VolumeTypes.LOSS_SCALE: np.dtype("float32"),
}, safe=True) +
Snapshot(every=50, output_filename='batch_{id}.hdf') +
PrintProfilingStats(every=50)
)
n = 500000
print("Training for", n, "iterations")
with gunpowder.build(batch_provider) as pipeline:
for i in range(n):
pipeline.request_batch(request)
print("Finished")
def test_fast_transform_no_recompute(self):
test_labels = ArrayKey("TEST_LABELS")
test_points = GraphKey("TEST_POINTS")
test_raster = ArrayKey("TEST_RASTER")
fast_pipeline = (DensePointTestSource3D() + ElasticAugment(
[10, 10, 10],
[0.1, 0.1, 0.1],
[0, 2.0 * math.pi],
use_fast_points_transform=True,
recompute_missing_points=False,
) + RasterizeGraph(
test_points,
test_raster,
settings=RasterizationSettings(radius=2, mode="peak"),
))
reference_pipeline = (
DensePointTestSource3D() +
ElasticAugment([10, 10, 10], [0.1, 0.1, 0.1], [0, 2.0 * math.pi]) +
RasterizeGraph(
test_points,
test_raster,
settings=RasterizationSettings(radius=2, mode="peak"),
))
timings = []
for i in range(5):
points_fast = {}
points_reference = {}
# seed chosen specifically to make this test fail
seed = i + 15
with build(fast_pipeline):
request_roi = Roi((0, 0, 0), (40, 40, 40))
request = BatchRequest(random_seed=seed)
request[test_labels] = ArraySpec(roi=request_roi)
request[test_points] = GraphSpec(roi=request_roi)
request[test_raster] = ArraySpec(roi=request_roi)
t1_fast = time.time()
batch = fast_pipeline.request_batch(request)
t2_fast = time.time()
points_fast = {
node.id: node
for node in batch[test_points].nodes
}
with build(reference_pipeline):
request_roi = Roi((0, 0, 0), (40, 40, 40))
request = BatchRequest(random_seed=seed)
request[test_labels] = ArraySpec(roi=request_roi)
request[test_points] = GraphSpec(roi=request_roi)
request[test_raster] = ArraySpec(roi=request_roi)
t1_ref = time.time()
batch = reference_pipeline.request_batch(request)
t2_ref = time.time()
points_reference = {
node.id: node
for node in batch[test_points].nodes
}
timings.append((t2_fast - t1_fast, t2_ref - t1_ref))
diffs = []
missing = 0
for point_id, point in points_reference.items():
if point_id not in points_fast:
missing += 1
continue
diff = point.location - points_fast[point_id].location
diffs.append(tuple(diff))
self.assertAlmostEqual(
np.linalg.norm(diff),
0,
delta=1,
msg=
"fast transform returned location {} but expected {} for point {}"
.format(point.location, points_fast[point_id].location,
point_id),
)
t_fast, t_ref = [np.mean(x) for x in zip(*timings)]
self.assertLess(t_fast, t_ref)
self.assertGreater(missing, 0)
def test_random_seed(self):
test_labels = ArrayKey('TEST_LABELS')
test_points = GraphKey('TEST_POINTS')
test_raster = ArrayKey('TEST_RASTER')
pipeline = (
PointTestSource3D() + ElasticAugment(
[10, 10, 10],
[0.1, 0.1, 0.1],
# [0, 0, 0], # no jitter
[0, 2.0 * math.pi]) + # rotate randomly
# [math.pi/4, math.pi/4]) + # rotate by 45 deg
# [0, 0]) + # no rotation
RasterizeGraph(test_points,
test_raster,
settings=RasterizationSettings(radius=2,
mode='peak')) +
Snapshot(
{
test_labels: 'volumes/labels',
test_raster: 'volumes/raster'
},
dataset_dtypes={test_raster: np.float32},
output_dir=self.path_to(),
output_filename='elastic_augment_test{id}-{iteration}.hdf'))
batch_points = []
for _ in range(5):
with build(pipeline):
request_roi = Roi((-20, -20, -20), (40, 40, 40))
request = BatchRequest(random_seed=10)
request[test_labels] = ArraySpec(roi=request_roi)
request[test_points] = GraphSpec(roi=request_roi)
request[test_raster] = ArraySpec(roi=request_roi)
batch = pipeline.request_batch(request)
labels = batch[test_labels]
points = batch[test_points]
batch_points.append(
tuple((node.id, tuple(node.location))
for node in points.nodes))
# the point at (0, 0, 0) should not have moved
data = {node.id: node for node in points.nodes}
self.assertTrue(0 in data)
labels_data_roi = (
labels.spec.roi -
labels.spec.roi.get_begin()) / labels.spec.voxel_size
# points should have moved together with the voxels
for node in points.nodes:
loc = node.location - labels.spec.roi.get_begin()
loc = loc / labels.spec.voxel_size
loc = Coordinate(int(round(x)) for x in loc)
if labels_data_roi.contains(loc):
self.assertEqual(labels.data[loc], node.id)
for point_data in zip(*batch_points):
self.assertEqual(len(set(point_data)), 1)
def test_3d_basics(self):
test_labels = ArrayKey("TEST_LABELS")
test_graph = GraphKey("TEST_GRAPH")
test_raster = ArrayKey("TEST_RASTER")
pipeline = (
GraphTestSource3D() + ElasticAugment(
[10, 10, 10],
[0.1, 0.1, 0.1],
# [0, 0, 0], # no jitter
[0, 2.0 * math.pi],
) + # rotate randomly
# [math.pi/4, math.pi/4]) + # rotate by 45 deg
# [0, 0]) + # no rotation
RasterizePoints(
test_graph,
test_raster,
settings=RasterizationSettings(radius=2, mode="peak"),
) + Snapshot(
{
test_labels: "volumes/labels",
test_raster: "volumes/raster"
},
dataset_dtypes={test_raster: np.float32},
output_dir=self.path_to(),
output_filename="elastic_augment_test{id}-{iteration}.hdf",
))
for _ in range(5):
with build(pipeline):
request_roi = Roi((-20, -20, -20), (40, 40, 40))
request = BatchRequest()
request[test_labels] = ArraySpec(roi=request_roi)
request[test_graph] = GraphSpec(roi=request_roi)
request[test_raster] = ArraySpec(roi=request_roi)
batch = pipeline.request_batch(request)
labels = batch[test_labels]
graph = batch[test_graph]
# the node at (0, 0, 0) should not have moved
# The node at (0,0,0) seems to have moved
# self.assertIn(
# Node(id=0, location=np.array([0, 0, 0])), list(graph.nodes)
# )
self.assertIn(0, [v.id for v in graph.nodes])
labels_data_roi = (
labels.spec.roi -
labels.spec.roi.get_begin()) / labels.spec.voxel_size
# graph should have moved together with the voxels
for node in graph.nodes:
loc = node.location - labels.spec.roi.get_begin()
loc = loc / labels.spec.voxel_size
loc = Coordinate(int(round(x)) for x in loc)
if labels_data_roi.contains(loc):
self.assertEqual(labels.data[loc], node.id)
def visualize_original_elastic_augment(args=None):
data_providers = []
data_dir = '/groups/saalfeld/home/hanslovskyp/experiments/quasi-isotropic/data'
file_pattern = 'sample_A_padded_20160501-2-additional-sections-fixed-offset.h5'
file_pattern = 'sample_B_padded_20160501-2-additional-sections-fixed-offset.h5'
file_pattern = 'sample_C_padded_20160501-2-additional-sections-fixed-offset.h5'
for data in glob.glob(os.path.join(data_dir, file_pattern)):
h5_source = Hdf5Source(data, datasets={
RAW: 'volumes/raw',
})
data_providers.append(h5_source)
input_resolution = (360, 36, 36)
output_resolution = Coordinate((120, 108, 108))
offset = (13640, 10932, 10932)
output_shape = Coordinate((12, 100, 100)) * output_resolution
output_offset = (13320 + 3600, 32796 + 36 + 10800, 32796 + 36 + 10800)
overhang = Coordinate((360, 108, 108)) * 16
input_shape = output_shape + overhang * 2
input_offset = Coordinate(output_offset) - overhang
output_roi = Roi(offset=output_offset, shape=output_shape)
input_roi = Roi(offset=input_offset, shape=input_shape)
augmentations = (
# SimpleAugment(transpose_only=[1,2]),
# ElasticAugmentNonMatchingVoxelSize(control_point_spacing=(1, 1, 1), jitter_sigma=(0.0, 3.0, 3.0), rotation_interval=(0, np.pi/2.0)),
ElasticAugment(
control_point_spacing=(4, 40, 40),
#jitter_sigma=(0, 1 * 2 * 36, 0 * 36),
jitter_sigma=(0, 2, 2),
rotation_interval=(2 * np.pi / 8, 0 * 2 * np.pi),
subsample=1), )
batch, snapshot = fuse.util.run_augmentations(
data_providers=data_providers,
roi=lambda key: input_roi.copy(),
augmentations=augmentations,
keys=(RAW, ),
voxel_size=lambda key: input_resolution)
args = get_parser().parse_args() if args is None else args
jnius_config.add_options('-Xmx{}'.format(args.max_heap_size))
import payntera.jfx
import imglyb
from jnius import autoclass
payntera.jfx.init_platform()
PainteraBaseView = autoclass(
'org.janelia.saalfeldlab.paintera.PainteraBaseView')
viewer = PainteraBaseView.defaultView()
pbv = viewer.baseView
scene, stage = payntera.jfx.start_stage(viewer.paneWithStatus.getPane())
payntera.jfx.invoke_on_jfx_application_thread(
lambda: pbv.orthogonalViews().setScreenScales([0.3, 0.1, 0.03]))
keys_to_show = (RAW, )
snapshot_states = add_to_viewer(
snapshot,
keys=keys_to_show,
name=lambda key: '%s-snapshot' % key.identifier)
states = add_to_viewer(batch, keys=keys_to_show)
viewer.keyTracker.installInto(scene)
scene.addEventFilter(
autoclass('javafx.scene.input.MouseEvent').ANY, viewer.mouseTracker)
while stage.isShowing():
time.sleep(0.1)
output_offset = (13320 + 3600, 32796 + 36 + 10800, 32796 + 36 + 10800)
overhang = Coordinate((360, 108, 108)) * 16
input_shape = output_shape + overhang * 2
input_offset = Coordinate(output_offset) - overhang
output_roi = Roi(offset=output_offset, shape=output_shape)
input_roi = Roi(offset=input_offset, shape=input_shape)
augmentations = (
# SimpleAugment(transpose_only=[1,2]),
# ElasticAugmentNonMatchingVoxelSize(control_point_spacing=(1, 1, 1), jitter_sigma=(0.0, 3.0, 3.0), rotation_interval=(0, np.pi/2.0)),
ElasticAugment(
control_point_spacing=(4, 40, 40),
#jitter_sigma=(0, 1 * 2 * 36, 0 * 36),
jitter_sigma=(0, 2, 2),
rotation_interval=(2 * np.pi / 8, 0 * 2 * np.pi),
subsample=1), )
batch, snapshot = fuse.util.run_augmentations(
data_providers=data_providers,
roi=lambda key: input_roi.copy(),
augmentations=augmentations,
keys=(RAW, ),
voxel_size=lambda key: input_resolution)
jnius_config.add_options('-Xmx{}'.format(args.max_heap_size))
import payntera.jfx
import imglyb
from jnius import autoclass
