- updated to
imagej-tensorflow 1.1.4
- native JNIs can now be switched via
Edit > Options > TensorFlow...
GenericCoreNetwork: remove menu path
- Adds
GenericTableNetwork
- add
scijava-plugins-io-table dependency
- rename
DefaultOutputProcessor to DatasetOutputProcessor
- add
TableOutputProcessor
- move everything but the
output Parameter and the initialization of the OutputProcessor from GenericNetwork to GenericCoreNetwork
GenericCoreNetwork: protected abstract initOutputProcessor method
GenericCoreNetwork: protected abstract computeOutput method
- make
GenericNetwork extending GenericCoreNetwork with Dataset output parameter
- create
GenericTableNetwork extending GenericCoreNetwork with GenericTable output parameter
OutputProcessor is typed now, changed signature of run
ImageTensor: add tilingAllowed attribute
ImageTensor: make a few methods private
POM: pom-scijava 23.2.0 → 27.0.1
PercentileNormalizer: Fix normalization for images with the same value for all pixels
- Images with the same value were normalized to images of value
NaN
- now they return an image of value
0.0
- don't load
meta.json of model since we do not use it yet
- !!!
groupId and artifactId changed !!!
<groupId>de.csbdresden</groupId>
<artifactId>csbdeep</artifactId>
- package names changed in the same fashion, e.g. usage in Python script:
from de.csbdresden.csbdeep.commands import GenericNetwork
- Commands now return
Dataset instead of List<Dataset> as output:
@Parameter(type = ItemIO.OUTPUT)
protected Dataset output;
- updated to TensorFlow CPU Version 1.12.0
- Adds parameter to hide process dialog (
showProcessDialog)
- fixes for input with singleton dimensions
- removes singleton dimensions of output
- removes unused
batchAxis parameter
- hopefully fixing thread issues with swing actions
- prints duration time of plugin run
- tiling improved
- see github commits for details
- manageable OOM throws now WARNING instead of ERROR