def volumina_n_layer(data, labels = None):
app = QApplication(sys.argv)
import volumina
from volumina.api import Viewer
v = Viewer ()
v.title = " Volumina Demo "
v.showMaximized ()
for ind, d in enumerate(data):
layer_name = "layer_" + str(ind)
if labels is not None:
layer_name = labels[ind]
# get data type of the elements d, to determine
# if we use a grayscale overlay (float32) or a randomcolors overlay (uint) for labels
data_type = d.dtype
if data_type == np.float32 or data_type == np.float64:
v.addGrayscaleLayer(d , name = layer_name)
else:
v.addRandomColorsLayer(d.astype(np.uint32), name = layer_name)
app.exec_()
def volumina_flexible_layer(data, layer_types, labels=None):
assert len(layer_types) == len(data)
app = QApplication (sys.argv)
import volumina
from volumina.api import Viewer
v = Viewer ()
v.title = " Volumina Demo "
v.showMaximized ()
for i, d in enumerate(data):
layer_name = "layer_" + str(i)
if labels is not None:
layer_name = labels[i]
# get data type of the elements d, to determine
# if we use a grayscale overlay (float32) or a randomcolors overlay (uint) for labels
data_type = d.dtype
if layer_types[i] == 'Grayscale':
v.addGrayscaleLayer(d , name = layer_name)
elif layer_types[i] == 'RandomColors':
v.addRandomColorsLayer(d.astype(np.uint32), name=layer_name)
elif layer_types[i] == 'Red':
v.addAlphaModulatedLayer(d , name=layer_name, tintColor=QColor(255,0,0))
elif layer_types[i] == 'Green':
v.addAlphaModulatedLayer(d , name=layer_name, tintColor=QColor(0,255,0))
elif layer_types[i] == 'Blue':
v.addAlphaModulatedLayer(d , name=layer_name, tintColor=QColor(0,0,255))
else:
raise KeyError("Invalid Layer Type, %s!" % layer_types[i])
app.exec_()
def view(data, labels=None, layer_types=None):
"""
"""
if labels is not None:
assert len(labels) == len(data)
if layer_types is not None:
assert len(layer_types) == len(data)
assert all(ltype in ('Grayscale', 'RandomColors', 'Red', 'Green', 'Blue')
for ltype in layer_types)
app = QApplication(sys.argv)
from volumina.api import Viewer
v = Viewer()
v.title = "Volumina Viewer"
v.showMaximized()
for i, d in enumerate(data):
layer_name = layer_name = "layer_" + str(i) if labels is None else labels[i]
if layer_types is None:
_dtype_to_layer(v, d, layer_name)
else:
_name_to_layer(v, d, layer_types[i], layer_name)
app.exec_()
def addHocViewer(grayData=None, segData=None, title="viewer", visu=True):
if visu:
app = QApp.Instance().app
v = Viewer()
if grayData is not None:
for name in grayData.keys():
data = grayData[name]
if _hasVigra:
if isinstance(data, vigra.arraytypes.VigraArray):
v.addGrayscaleLayer(data.view(numpy.ndarray),
name=name)
else:
v.addGrayscaleLayer(data, name=name)
else:
v.addGrayscaleLayer(data, name=name)
if segData is not None:
for name in segData.keys():
data = segData[name]
if _hasVigra:
if isinstance(data, vigra.arraytypes.VigraArray):
v.addColorTableLayer(data.view(numpy.ndarray),
name=name)
else:
v.addColorTableLayer(data, name=name)
else:
v.addGrayscaleLayer(data, name=name)
v.setWindowTitle(title)
v.showMaximized()
app.exec_()
def volumina_single_layer(data): app = QApplication (sys.argv) from volumina.api import Viewer v = Viewer () v.title = " Volumina Demo " v.showMaximized () v.addGrayscaleLayer (data , name =" raw data ") app . exec_ ()
def volumina_viewer():
app = QApplication(sys.argv)
v = Viewer()
v.editor.setInteractionMode("navigation")
yield v
try:
signal.signal(signal.SIGINT, lambda *_: app.quit())
app.exec_()
finally:
app.quit()
def streaming_n_layer(files, keys, labels=None, block_shape=[100, 100, 100]):
from volumina.api import Viewer
from volumina.pixelpipeline.datasources import LazyflowSource
from lazyflow.graph import Graph
from lazyflow.operators.ioOperators.opStreamingHdf5Reader import OpStreamingHdf5Reader
from lazyflow.operators import OpCompressedCache
app = QApplication(sys.argv)
v = Viewer()
graph = Graph()
def mkH5source(fname, gname):
h5file = h5py.File(fname)
dtype = h5file[gname].dtype
source = OpStreamingHdf5Reader(graph=graph)
source.Hdf5File.setValue(h5file)
source.InternalPath.setValue(gname)
op = OpCompressedCache(parent=None, graph=graph)
op.BlockShape.setValue(block_shape)
op.Input.connect(source.OutputImage)
return op.Output, dtype
#rawSource = mkH5source(data[0], keys[0])
#v.addGrayscaleLayer(rawSource, name = 'raw')
for i, f in enumerate(files):
if labels is not None:
layer_name = labels[i]
else:
layer_name = "layer_%i" % (i)
source, dtype = mkH5source(f, keys[i])
if np.dtype(dtype) in (np.dtype('uint8'), np.dtype('float32'),
np.dtype('float64')):
v.addGrayscaleLayer(source, name=layer_name)
else:
v.addRandomColorsLayer(source, name=layer_name)
v.setWindowTitle("Streaming Viewer")
v.showNormal()
app.exec_()
def volumina_double_layer(data, overlay): # get data type of the elements of overlay, to determine # if we use a grayscale overlay (float32) or a randomcolors overlay (uint) for labels mask = [] for i in range( len(overlay.shape) ): mask.append(0) mask = tuple(mask) data_type = type(overlay[mask]) app = QApplication (sys.argv) from volumina.api import Viewer v = Viewer () v.title = " Volumina Demo " v.showMaximized () v.addGrayscaleLayer(data , name = " raw data ") if data_type == np.float32: v.addGrayscaleLayer(overlay , name = " overlay ") else: v.addRandomColorsLayer(overlay, name = " overlay ") app . exec_ ()
if len(args) != 1:
parser.error("no hdf5 dataset supplied")
x = args[0].find(".h5")
fname = args[0][:x + 3]
gname = args[0][x + 4:]
# load data
f = h5py.File(fname, "r")
raw = f[gname].value
assert raw.ndim == 3
assert raw.dtype == numpy.uint8
f.close()
app = QApplication([])
v = Viewer()
direct = True
raw.shape = (1, ) + raw.shape + (1, )
l1 = v.addGrayscaleLayer(raw, name="raw", direct=True)
l1.visible = direct
colortable = [
QColor(0, 0, 0, 0).rgba(),
QColor(255, 0, 0).rgba(),
QColor(0, 255, 0).rgba(),
QColor(0, 0, 255).rgba()
]
s = ((raw // 64)).astype(numpy.uint8)
a = np.zeros(500 * 500).reshape(500, 500).astype(np.uint8)
ii = np.random.randint(0, 500, 1)
jj = np.random.randint(0, 500, 1)
a[ii, jj] = 1
a = vigra.filters.discDilation(a, radius=20)
array[:] = a.reshape(shape).view(np.ndarray) * 255
op.Input.setDirty()
do()
cron.timeout.connect(do)
ds = createDataSource(op.Output)
layer = ColortableLayer(ds, jet())
mainwin = Viewer()
mainwin.layerstack.append(layer)
mainwin.dataShape = (1, h, w, 1, 1)
print(mainwin.centralWidget())
BoxContr = BoxController(mainwin.editor, op.Output, boxListModel)
BoxInt = BoxInterpreter(mainwin.editor.navInterpret, mainwin.editor.posModel, BoxContr, mainwin.centralWidget())
mainwin.editor.setNavigationInterpreter(BoxInt)
# boxListModel.boxRemoved.connect(BoxContr.deleteItem)
LV.show()
mainwin.show()
app.exec_()
def showStuff(raw_name,
pred_viewer1,
pred_viewer2,
cutout_name,
one_extra=None):
# display the raw and annotations for cremi challenge data
raw = vigra.impex.readHDF5(indir + datasets[raw_name], "data", order='C')
# raw_old = vigra.readHDF5(indir+datasets["raw_bad"], "data", order = 'C')
defect_prediction_128 = vigra.impex.readHDF5(indir +
datasets[pred_viewer2],
"data",
order='C')
defect_prediction_150 = vigra.impex.readHDF5(indir +
datasets[pred_viewer1],
"data",
order='C')
cutout_from_150_pred = vigra.impex.readHDF5(indir + datasets[cutout_name],
"data",
order='C')
####################################################################################################################
# only used for fast testing stuff
#change_one = vigra.readHDF5(indir+datasets["segmentation_on_equalized_image"], "data", order = 'C')
#pdb.set_trace()
#defect_prediction_150[1,:,:] = change_one[0,:,:,0]
####################################################################################################################
# defect_prediction_150 = gt[..., 0]
cutout = numpy.asarray(cutout_from_150_pred)
rawdata = numpy.asarray(raw)
# rawdata_old = numpy.asarray(raw_old)
# op5ify
# shape5d = rawdata.shape
shape5d = (1, ) + rawdata.shape + (1, )
print shape5d, rawdata.shape, rawdata.dtype
app = QApplication([])
v = Viewer()
direct = False
# layer for raw data
rawdata = numpy.reshape(rawdata, shape5d)
rawsource = ArraySource(rawdata)
v.dataShape = shape5d
lraw = GrayscaleLayer(rawsource, direct=direct)
lraw.visible = True
lraw.name = "raw"
v.layerstack.append(lraw)
# layer for cutout regions from raw data
cutout = numpy.reshape(cutout, shape5d)
cutoutsource = ArraySource(cutout)
lcutout = GrayscaleLayer(cutoutsource, direct=direct)
lcutout.visible = False
lcutout.name = "cut_out"
v.layerstack.append(lcutout)
# layer for first prediction result
defect_prediction_128 = numpy.reshape(defect_prediction_128, shape5d)
synsource = ArraySource(defect_prediction_128)
ct = create_random_16bit()
ct[0] = 0
lsyn = ColortableLayer(synsource, ct)
lsyn.name = pred_viewer2
lsyn.visible = False
v.layerstack.append(lsyn)
# layer for second prediction result
segm = numpy.reshape(defect_prediction_150, shape5d)
segsource = ArraySource(segm)
ct = create_random_16bit()
ct[0] = 0
lseg = ColortableLayer(segsource, ct)
lseg.name = pred_viewer1
lseg.visible = False
v.layerstack.append(lseg)
if one_extra is None:
v.showMaximized()
app.exec_()
if one_extra is not None:
# layer for third prediction result
extra_prediction = vigra.readHDF5(indir + datasets[one_extra],
"data",
order='C')
extra_pred_reshaped = numpy.reshape(extra_prediction, shape5d)
segsource = ArraySource(extra_pred_reshaped)
ct = create_random_16bit()
ct[0] = 0
# ct = create_default_16bit()
lseg = ColortableLayer(segsource, ct)
lseg.name = one_extra
lseg.visible = False
v.layerstack.append(lseg)
v.showMaximized()
app.exec_()
