Releases: BiaPyX/BiaPy
Releases · BiaPyX/BiaPy
Version 3.3.9
Version 3.3.8
Version 3.3.7
Changes:
- Now
DATA.TEST.ARGMAX_TO_OUTPUT
defaults toTrue
- Add multi-head in instance segmentation workflow to identify the class of each instance
Fixes:
- Change rotate to scipy so it can be used with 3D images
- Change TTA to allow multiple heads as output of the network
- Fix minor error in BMZ export
- Fix edge case when using
DATA.REFLECT_TO_COMPLETE_SHAPE
- Reduce memory comsuption in merge functions
Full Changelog: v3.3.6...v3.3.7
Version 3.3.6
Changes:
AUGMENTOR.RANDOM_ROT
andAUGMENTOR.ROT90
now are implemented in BiaPy and not done through imgaug.- Add
TRAIN.VERBOSE
to visualize more or less info during each batch process print
Fixes:
- Fix 4 dims length Zarr data creation during
TEST.BY_CHUNKS
. - Change slightly custom architectures (
MODEL.SOURCE
==biapy
) so they can be converted into TorchScript viatorch.jit.script()
to create BMZ package. - Fix U-Net like models for SR to depend on
PROBLEM.SUPER_RESOLUTION.UPSCALING
factor and allowMODEL.Z_DOWN
in super-resolution workflow - Limit number of workers per GPU for safety
- Fix crappify issues for SSL
Version 3.3.5
Fix patch:
- Rename
PROBLEM.NUM_CPUS
toPROBLEM.NUM_WORKERS
to clarify its usage. - Speed up SSL workflow
Version 3.3.4
Changes:
- Set
TEST.DET_EXCLUDE_BORDER
toFalse
by default. - Add
TEST.DET_PEAK_LOCAL_MAX_MIN_DISTANCE
. - 3 int tuple for TEST.RESOLUTION in instance segmentation if
TEST.ANALIZE_2D_IMGS_AS_3D_STACK
. - Prevent usage of EfficientNet architectures for 3D.
- Add
PROBLEM.INSTANCE_SEG.WATERSHED_BY_2D_SLICE
.
Fix:
- Prevent creating multiple processes to manage data if low samples are available.
- Solve EfficientNet issue with biapy backend as discussed here.
- Bug in instance seg when no labels are provided.
- Disable aug sample image generation if DA is disabled.
- Fix SSL bug during training due to recent changes.
Version 3.3.3
Fixes:
- Change
DATA.PREPROCESS.*.ACTIVATE
toDATA.PREPROCESS.*.ENABLE
as the rest of the variables in all the files (changed only in config.py by error). - Separate per_image, full_image and as_3D_stack instance files in different folders.
- Separate instance segmentation metrics when multiple choices are selected. Before full_image and per_image metrics were mixed.
- Simplify inference by setting as default patch/merge reconstruction of the prediction. This implied to remove
TEST.STATS
and leave onlyFULL_IMG
to be optional. TEST.FULL_IMG
toFalse
by default.
Version 3.3.2
Quick patch to fix some issues:
- Move
sys.exit()
call tomain.py
to prevent errors inside jupyter notebooks - Fix issue during BMZ export in classification
- Rename
DATA.PREPROCESSING.*.ACTIVATE
toENABLE
as in other variables. - Remove
DATA.PREPROCESS.MEDIAN_BLUR.FOOTPRINT
as it is a Numpy array and it can not be declared through YACS
Version 3.3.1
Quick patch to fix some issues:
- Fix
FORCE_RGB
variable usage in classification - Adapt skimage's
relabel_sequential()
to be as the old function we were using so the matching metrics process doesn't get stuck anymore.
Version 3.3.0
General changes
Major
- Separate instance filtering and statistical measurements with
TEST.POST_PROCESSING.MEASURE_PROPERTIES
andTEST.POST_PROCESSING.MEASURE_PROPERTIES.REMOVE_BY_PROPERTIES
- Add sphericity (3D), perimeter/surface area (2D/3D) and elongation (2D) calculations using the same formulas as described in MorphoLibJ
- Multi-GPU prediction by chunks (Zarr/H5):
- Add versatile axis order
- Fix some overlap errors
- Add data preprocessing options:
- Resize
- Gaussian blur
- Median blur
- Histogram matching
- Contrast Limited Adaptive Histogram Equalization (CLAHE)
- Canny or edge detection (only 2D - grayscale or RGB)
- Change BiaPy into a class so we can call functions individually (e.g. BMZ model exportation)
- Detection:
- Add overlap in detection during multi-GPU prediction by chunks
- Now point coords work in global position
Minor
- Add
TEST.DET_EXCLUDE_BORDER
option
Bugs fixed:
- 2D test time augmentation bug with
MODEL.N_CLASSES
solved - Fix bug when
TEST.BY_CHUNKS
selected usingTEST.BY_CHUNKS.INPUT_IMG_AXES_ORDER
of len 4. - Avoid dividing with zero during instance stats