nt

When saving a project snapshot I do not want to touch the original project file. It should be clear by the name of the option that only a new snapshot will be written out but nothing in the original project file will be touched.

Rename to "save snapshot" or "take snapshot" or something like that

When I want to change the opacity of a layer, or toggle its visibility, then in most cases I need two clicks, and thus one too many: first click to set focus on a layer, second click to achieve what I want. Can we set focus AND perform action by a single click? If not, we'll need up / down / remove icons on every layer.

... is causing the ilastik window to freeze forever. I mean, adding a completely new label to the label widget, not adding a new stroke of the old label.

The sigma values shown in the feature selection applet are actually window sizes, not sigmas. Real sigmas should be shown.

The widgets in the igms repo should live inside the shellwidgets module of the ilastikshell project. We don't need these two separated.

Hi,

When I want to run batch processing for a 1GB input file, select about 15 features, and select "save to directory", I am always getting a segmentation fault after a while. (I haven't tried that out with a smaller dataset or considerably less features).
This happens on my Linux Mint workstation and Ubuntu 12.04 laptop.

Let me know if you need more details.
Backtrace:

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7fffe53f8700 (LWP 24840)]
0x00007ffff0130111 in pyx_tp_dealloc_4h5py_8_objects_ObjectID (o=
<traceback at remote 0x5be8d88>) at h5py/_objects.c:4570
4570 h5py/_objects.c: No such file or directory.
(gdb) bt
0 0x00007ffff0130111 in __pyx_tp_dealloc_4h5py_8_objects_ObjectID (o=
<traceback at remote 0x5be8d88>) at h5py/_objects.c:4570
1 0x00007fffebe64718 in __pyx_pf_4h5py_3h5d_9DatasetID_5shape___get (
__pyx_v_self=) at h5py/h5d.c:2040
2 __pyx_getprop_4h5py_3h5d_9DatasetID_shape (o=,
x=) at h5py/h5d.c:3590
3 0x00000000004e6038 in ?? ()
4 0x00000000004ee36a in _PyObject_GenericGetAttrWithDict ()
5 0x0000000000497f4e in PyEval_EvalFrameEx ()
6 0x000000000049f1c0 in PyEval_EvalCodeEx ()
7 0x00000000004a8960 in ?? ()
8 0x00000000004e9f36 in PyObject_Call ()
9 0x0000000000552895 in PyObject_CallFunction ()
10 0x00000000004ee36a in _PyObject_GenericGetAttrWithDict ()
11 0x0000000000497f4e in PyEval_EvalFrameEx ()
12 0x000000000049f1c0 in PyEval_EvalCodeEx ()
13 0x00000000004a8960 in ?? ()
14 0x00000000004e9f36 in PyObject_Call ()
15 0x00000000004ec11a in ?? ()
16 0x00000000004e9f36 in PyObject_Call ()
17 0x0000000000479f97 in ?? ()
18 0x0000000000499171 in PyEval_EvalFrameEx ()
19 0x0000000000498602 in PyEval_EvalFrameEx ()
--Type to continue, or q to quit---
20 0x000000000049f1c0 in PyEval_EvalCodeEx ()
21 0x00000000004a8a92 in ?? ()
22 0x00000000004e9f36 in PyObject_Call ()
23 0x00000000004ec11a in ?? ()
24 0x00000000004e9f36 in PyObject_Call ()
25 0x00000000004eb62e in ?? ()
26 0x00000000004e9f36 in PyObject_Call ()
27 0x0000000000499bc0 in PyEval_EvalFrameEx ()
28 0x0000000000498602 in PyEval_EvalFrameEx ()
29 0x000000000049f1c0 in PyEval_EvalCodeEx ()
30 0x00000000004983b8 in PyEval_EvalFrameEx ()
31 0x0000000000498602 in PyEval_EvalFrameEx ()
32 0x000000000049f1c0 in PyEval_EvalCodeEx ()
33 0x00000000004a8a92 in ?? ()
34 0x00000000004e9f36 in PyObject_Call ()
35 0x00000000004ec11a in ?? ()
36 0x00000000004e9f36 in PyObject_Call ()
37 0x00000000004eb62e in ?? ()
38 0x00000000004e9f36 in PyObject_Call ()
39 0x0000000000499bc0 in PyEval_EvalFrameEx ()
40 0x0000000000498602 in PyEval_EvalFrameEx ()
41 0x000000000049f1c0 in PyEval_EvalCodeEx ()
42 0x00000000004983b8 in PyEval_EvalFrameEx ()
--Type to continue, or q to quit---
43 0x000000000049f1c0 in PyEval_EvalCodeEx ()
44 0x00000000004a8a92 in ?? ()
45 0x00000000004e9f36 in PyObject_Call ()
46 0x00000000004ec11a in ?? ()
47 0x00000000004e9f36 in PyObject_Call ()
48 0x000000000047b6b7 in ?? ()
49 0x00000000004e9f36 in PyObject_Call ()
50 0x0000000000499bc0 in PyEval_EvalFrameEx ()
51 0x000000000049f1c0 in PyEval_EvalCodeEx ()
52 0x00000000004a8960 in ?? ()
53 0x00000000004e9f36 in PyObject_Call ()
54 0x00000000004ec11a in ?? ()
55 0x00000000004e9f36 in PyObject_Call ()
56 0x00000000004ea396 in PyEval_CallObjectWithKeywords ()
57 0x00007fffea7aea1d in g_initialstub (mark=)
at greenlet.c:804
58 0x00007fffea7ae368 in g_switch (target=0x7fff5043e260, args=(), kwargs=0x0)
at greenlet.c:576
59 0x00007fffea7aeef6 in green_switch (self=,
args=, kwargs=) at greenlet.c:1092
60 0x0000000000497ea4 in PyEval_EvalFrameEx ()
61 0x0000000000498602 in PyEval_EvalFrameEx ()
62 0x0000000000498602 in PyEval_EvalFrameEx ()
--Type to continue, or q to quit---
63 0x000000000049f1c0 in PyEval_EvalCodeEx ()
64 0x00000000004a8960 in ?? ()
65 0x00000000004e9f36 in PyObject_Call ()
66 0x00000000004ec11a in ?? ()
67 0x00000000004e9f36 in PyObject_Call ()
68 0x00000000004ea396 in PyEval_CallObjectWithKeywords ()
69 0x00000000004b0bc8 in ?? ()
70 0x00007ffff7bc4e9a in start_thread (arg=0x7fffe53f8700)
at pthread_create.c:308
71 0x00007ffff69b5dbd in clone ()
at ../sysdeps/unix/sysv/linux/x86_64/clone.S:112
72 0x0000000000000000 in ?? ()

The GUI should wait until the user actually opens a new project before cleaning up the current project.

Right now, hitting "File->Open Project", then "Cancel", does not allow the user to resume working on the current project.

In the Pixel Classification Training Applet, when I start to label with one brush stroke, this stroke is (partly) forgotten as soon as the label is rendered. This seems to happen only for the first stroke(s) of each label.

The feature selection dialog should be modal and not just floating around somewhere in the background. I opened and forget about it and when shift+tabbing between the windows I got either the dialog or the main window.

Zoom out such that several tiles are in field of view. Switch off all layers except for a segmentation layer and the raw data.

Add annotations within a single tile only.

Then only the segmentation in that single tile is updated.

We need keyboard shortcuts for the Project menu and the layer list controls. Some applet drawer controls (e.g. "Live Prediction") should have keyboard shortcuts, too.

We could have a neutral light gray everywhere, or even a gradient in the orthoviews (woohoo)

When I select all features at a given scale, the "Difference of Gaussians" and "Laplacian of Gaussian" are visually indiscernible. Anyway, do we want both?

to accommodate all lines of the menu. When I make the window larger, I retain the scrollbar on the RHS, the rest of the contents does not unfold.

Anyway, we need something prettier at some stage.

Add the old feature categories to the feature selection dialog so ilastik 0.5 users know what to select.

I load a volume and train.
Then I load additional volumes of different sizes.
I delete all original label classes (by pressing the red cross next to the label classes).
When switching between volumes, I no longer see the raw data.
When adding new labels, their colors are not correct (it does not start with red / green)
I can not add new annotations (which anyway doesn't make sense, given that the raw data is not visible)

• Rename ilastik to ilastik 0.5
• Rename applet-workflows to ilastik 0.6
• Create ilastik-meta, which tracks the working revisions of lazyflow, ilastik, and volumina.

Implement in LayerViewerGui base class.

Right now, closing the main window doesn't call the same callback that choosing "Quit" does.

In the past people where confused about "Color" features especially when dealing with grayscale data. We should rename them to "Intensity" (or maybe "Lightness" or "Value")

when I do the following:

• select all features at scale 11.
• ok
• deselect Gaussian smoothing at scale 11, add Gaussian smoothing at scale 3.

Since I (and probably other users) do not always know which repo is responsible for a bug, we need to know where .6 bugs should be reported.

Also, make sure that the user can switch back and forth between relative and absolute paths.

which of the modes is activated

Apparently some users want to be capable of viewing project files in directories they have read-only access to. Currently, ilastik will simply fail to open the project (we assume we'll want to write to the file eventually). Instead, we should open the file anyway, and fail if the user tries to do save any changes to it.

User shortcut settings should be saved to to the hard disk and loaded on startup.

In the FeatureSelection GUI, the feature table widget allows the user to edit the sigma column headings. The current version of the applet uses hard-coded sigma options, so the user's edits will be ignored. Disable the edit feature for now to avoid user confusion.

We need a unit test for the Export Layer context menu item. The test should use a lazyflow slot to provide the data.

In addition, to make this functionality more discoverable, add a "save" button next to the up, down arrows. This will export the currently selected layer.

Here is the end of the stacktrace:

File "/home/akreshuk/lazyflow/lazyflow/graph.py", line 697, in call
result_op = self.operator.execute(self.slot, (), roi, destination)
File "/home/akreshuk/lazyflow/lazyflow/operators/obsolete/vigraOperators.py", line 1217, in execute
slicings=self.computeRequestSlicings()
File "/home/akreshuk/lazyflow/lazyflow/operators/obsolete/vigraOperators.py", line 1268, in computeRequestSlicings
shift = chunkShape * numpy.array([10,2,2,2,10])
ValueError: operands could not be broadcast together with shapes (4) (5)

If possible with little work, would be nice to have a switch to use smaller fonts there.

to allow loading multiple jpg, tiff or png images into a single 3D volume.

To reproduce:

1. New Project, add 5d data from hdf5
2. Go to training drawer and click on the upper triangle of the timespinbox
   -> Jumps from zero to 2

This is not happening when the ImageView is maximized.

Bug appears with the following commit to volumina
3fe2832d2130703f462ef0268160b3a3d400eefc

The bug appears only in pixelclassification, but not in standalone volumina.

The commit looks unrelated, though. Could be a bug somehow caused by the new tooltip or shortcut mechanism

When I want to run pixel classification batch processing from the command line with the following command

python pixelClassificationWorkflowMainHeadless.py --project ~/projectFile.ilp --batch_export_dir ~/outputDir/ ~/inputFile.h5

I am getting the following error message:

INFO main: Opening project: '/home/user/projectFile.ilp'
INFO ilastik.shell.projectManager: Opening Project: /home/user/projectFile.ilp

ERROR 2012-10-12 20:22:22,062 pixelClassificationWorkflowMainHeadless 16795 140659294799616 Traceback (most recent call last):
File "pixelClassificationWorkflowMainHeadless.py", line 35, in main
runWorkflow(parsed_args)
File "pixelClassificationWorkflowMainHeadless.py", line 102, in runWorkflow
args.batch_output_dataset_name)
File "pixelClassificationWorkflowMainHeadless.py", line 163, in generateBatchPredictions
logger.info( "Exporting data to " + opBatchResults.OutputDataPath[0].value )
File "/home/user/ilastikenv/local/lib/python2.7/site-packages/lazyflow-0.0.0-py2.7.egg/lazyflow/graph.py", line 830, in value
temp = self[:].allocate().wait()
File "/home/user/ilastikenv/local/lib/python2.7/site-packages/lazyflow-0.0.0-py2.7.egg/lazyflow/request.py", line 561, in wait
self._execute()
File "/home/user/ilastikenv/local/lib/python2.7/site-packages/lazyflow-0.0.0-py2.7.egg/lazyflow/request.py", line 682, in _execute
self.result = self.function(**self.kwargs)
File "/home/user/ilastikenv/local/lib/python2.7/site-packages/lazyflow-0.0.0-py2.7.egg/lazyflow/graph.py", line 694, in call
result_op = self.operator.execute(self.slot, (), roi, destination)
File "/home/user/ilastikenv/local/lib/python2.7/site-packages/ilastik-0.6a-py2.7.egg/ilastik/applets/batchIo/opBatchIo.py", line 112, in execute
assert False # This slot is already set via setupOutputs
AssertionError

Seems like setupOutputs of opBatchIo is never called?

The error doesn't appear if I run the batch processing with the --generate_project_predictions flag.

Shortcuts in volumina are not coordinated through the applet-workflows shortcut manager, so it isn't clear which shortcuts are unavailable for applet controls without inspecting the VolumeEditorWidget source code.

When selecting a file with invalid datatype the "open file" silently closes and nothing happens. Show an error message to give the user feedback.

Add a setup.py to allow
"python setup.py install".

This way, we don't have to mess around with PYTHONPATH anymore.

indicated point appears in the field of view

I'm sometimes (in 1 out of ~20 cases) getting the following exception when running batch processing. This time, the process got locked at 90%.

http://pastebin.com/J5LGzRB2

Load two volumes
compute features
label and predict interactively on one volume
switch to the other and view a few predictions
switch back to the first volume
switch to feature app

Now there are, in addition to the original, and fully functional, feature layers, some more redundant instances of the original feature layers. They are non-functional and show up as blank in the orthoviews.

ilastik 0.5 had a feature to export the current (2-d) slicing view. The new ilastik should have it, too.