Releases: smistad/FAST
Releases · smistad/FAST
v4.2.1
Changes
- Fixed bug with relative paths in MovieStreamer
- Added loop option to MovieStreamer
- Added framerate option to ImageFileStreamer
- Fixed bug in UFFStreamer when having UFF data with 2 dimensions instead of 4
- Implemented scan conversion, envelope detection and log compression on GPU as separate process objects.
- Added settings for gain and dynamic range in UFFviewer
- Improved performance of UFFStreamer
- Fixed bug with pixman library being loaded from system instead of lib folder
- Added possibility of specifying text pipeline input data PipelineInputData
- Added the ImageTransposer PO for transposing 2D and 3D images
- Added the Interleave PO for interleaving multiple data streams into one output stream by interleaving the data objects from the input streams
Full Changelog: v4.2.0...v4.2.1
Release binaries and installers can be downloaded below by pressing "assets". Make sure you have the required dependencies installed on your system before using the release: see the installation instructions for Windows, Ubuntu Linux, macOS and Python.
v4.2.0
Changes
- Experimental Mac support for macOS >= 10.13 including OpenVINO and TensorFlow CPU inference.
- Updated Clarius Cast to version 8.6
- Fixed a serious memory leak in Image introduced in FAST v4
- Improved performance of image/segmentation pyramid rendering by removing an expensive OpenGL synchronization.
- Added a "Null" color to allow vertex renderer to use custom colors.
- Added the ImageAdd process object which takes two images performs pixel-wise summation as the output.
- Added the IntensityClipping process object which clips the pixel intensity of an image to a given range.
- Added the ImageCaster process object which can cast an image from one data type to another.
- Added the Start/StopTimer process objects for easy runtime measurements
- Added ImagePyramidLevelExtractor process object which can extract a single level from an ImagePyramid as an Image.
- Added RunUntilFinished process object.
- Added a TissueMicroArrayExtractor streamer to extract TMAs as images from an image pyramid.
- Implemented text pipeline input data mechanism
- Added possibility of setting fixed min and max intensity in IntensityNormalization.
- Added resample if needed feature in ImageFileExporter
- Added multi-channel support in ImageResampler
- Use a S-like grayscale curve in UltrasoundImageEnhancement instead of a linear curve
- Added more features to pyFAST: Volume renderers, Line plotter, Batch and Sequence (std::vector/list of Image and Tensor)
- Added window/level and opacity options to ImageRenderer.
- Refactored ComputationThread so it can be used in Python.
- Added plotting styles bright and dark to LinePlotter.
- Enabled ImageFileImporter to open normal tif/tiff files.
- Added possibility of specifying which channels to use in TensorToImage.
- Enabled TIFFImagePyramidExporter to also accept regular images, and not just image pyramids
- Many bug fixes and more documentation +++
Full Changelog: v4.1.0...v4.2.0
Release binaries and installers can be downloaded below by pressing "assets". Make sure you have the required dependencies installed on your system before using the release: see the installation instructions for Windows, Ubuntu Linux, macOS and Python.
v4.2.0-rc3
This is a release candidate for testing purposes.
Full Changelog: v4.1.0...v4.2.0-rc3
v4.1.0
Changelog
- Fixed deadlock issues in pyFAST when using OpenCV in PythonProcessObjects
- Simplified python wrapping of getOutputData/runAndGetOutputData/getNextFrame methods. Instead of runAndGetOutputImage/Mesh etc., one should use runAndGetOutputData which will return the correct type.
- Added examples of using pyFAST with PySide2 for making Qt GUI with FAST in Python
- Added a fix for synchronized rendering blocking execution when window is minimized
- Added V4 semantics to FlowNetwork PO
- Added possibility of setting lung seed point in LungSegmentation
- Use bundled OpenCL with FAST
- Improved input/output node specification in neural networks
- Fixed issue with creating an empty Mesh
- Fixed bug in DualViewWindow vertical/horizontal mode
- Fixed bug in liverVesselSegmentation example
- Added possibility of setting address and port in Clarius streaming example
- Disabled 3D image writes on nvidia windows due to instability with latest drivers (471.35)
- Added possibility of warmup rounds in runtime measurements system
Release binaries and installers can be downloaded below by pressing "assets". Make sure you have the required dependencies installed on your system before using the release: see the installation instructions for Windows, Ubuntu Linux, and Python.
v4.0.0
New release with several major changes thereby increasing major version to 4.
Major changes:
- New pipeline syntax: Use create() instead of New(), and connect() instead of setInputConnection()/getOutputPort(). Check out the documentation to learn the new syntax.
- New build system: Dependencies are built using a separate repo, thereby reducing build times of FAST considerably.
Changelog:
- Updated TensorRT engine to use TensorRT 8
- Updated Clarius Cast to 8.0.1 (Note clarius cast only works on ubuntu 20.04 now)
- More documentation, see https://fast.eriksmistad.no
- GaussianSmoothingFilter renamed to GaussianSmoothing
- AffineTransformation renamed to Transform
- Simplified SimpleDataObject. No need for access objects.
- Can now use keyword arguments in the create() function in python.
- Implemented border size for BoundingBoxRenderer
- Introduced SimpleWindow2D/3D DualViewWindow2D/3D to remove need for set2DMode/set3DMode
- Refactored RunLambda
- Implemented DataStream object for C++ and Python making it easier to iterate through a data stream
- 3D support for crop bottom/top in ImageCropper
- Improved exception handling in pyFAST
- Simplified streaming mode setting. The setting is set on the streamer only, not by Config.
- ++ bug fixes and refactoring
Release binaries and installers can be downloaded below by pressing "assets". Make sure you have the required dependencies installed on your system before using the release: see the installation instructions for Windows, Ubuntu Linux, and Python.
v3.3.0
Changes
- systemCheck executable for checking FAST and OpenCL
- Refactored large parts of the rendering setup, moving more computations to shaders and introduced the getOpenGLTextureAccess on Image object.
- New and improved documentation, see https://fast.eriksmistad.no
- Made it possible to run FAST on remote server without visualization capabilities (headless)
- TIFFImagePyramidImporter and TIFFImagePyramidExporter - For importing and exporting tiled pyramidal TIFFs
- ImagePyramidPatchImporter/Exporter - For importing and exporting image pyramids as series of png images
- SegmentationPyramidRenderer removed and merged with SegmentationRenderer
- Implemented overlap in patch wise processing pipelines
- Bug fixes related to image pyramids
- Border opacity in SegmentationRenderer
- Improved image pyramid rendering speed
- Horizontal line feature in LinePlotter
- Fixed bug in python bindings on windows with eigen<->numpy integer array conversion not working #144
- Fixed bug with access type not exported to python bindings
Release binaries and installers can be downloaded below by pressing "assets". Make sure you have the required dependencies installed on your system before using the release.
Please see the installation instructions for Windows, Ubuntu Linux, and Python.
v3.2.0
Changelog
- Working python bindings and wheel creation using SWIG. Use py_limited_api abi3 for python bindings which enables one single binary wheel to be used on different python 3 versions.
- Added several python examples.
- Updated TensorFlow to version 2.4.0, removed need for separate CUDA/CPU versions, and added support for loading models in SavedModel format. Note: TensorFlow engine is now built with CUDA 11.0 and cuDNN 8.
- Updated TensorRT to version 7.2 and added support for ONNX models. Note: TensorRT engine is now built with CUDA 11.0 and cuDNN 8.
- Updated OpenVINO to version 2021.1 and added support for ONNX models.
- Updated Clarius cast to version 7.3.0
- Updated RealSense to version 2.40
- Updated eigen library to work with latest MSVC
- Updated JKQTPlotter library to work with latest MSVC
- Fixed issue with Intel OpenCL not working with FAST on Linux
- Added FlowNetwork
- Created an UFF viewer executable
- Updated windows installer to add start menu items
- Added the PythonProcessObject
- Updated the neural network models in the test data. You should redownload the test data as the old models might not work.
- Extended ImagePyramid to be able to create custom ImagePyramids
- Added SegmentationPyramidRenderer to render super high resolution segmentations stored as a pyramid.
- Added zip library and Qt Network to enable download of test data from software. Added executable downloadTestData which can be used to download test data. All test data is now put in C:/ProgramData/FAST/data/ and /home//FAST/data.
- Bux fixes etc ++
Release binaries and installers can be downloaded below, by pressing "assets". Please make sure you have the required dependencies installed on your system before using the release.
v3.2.0-rc1
THIS IS A PRE-RELEASE FOR TESTING, SEE LATEST STABLE RELEASE BELOW
Changelog
- Working python bindings and wheel creation using SWIG. Use py_limited_api abi3 for python bindings which enables one single binary wheel to be used on different python 3 versions.
- Added several python examples.
- Updated TensorFlow to version 2.4.0, removed need for separate CUDA/CPU versions, and added support for loading models in SavedModel format.
- Updated TensorRT to version 7.2 and added support for onnx models
- Updated OpenVINO to version 2021.1
- Updated Clarius cast to version 7.3.0
- Updated RealSense to version 2.40
- Updated eigen library to work with latest MSVC
- Updated JKQTPlotter library to work with latest MSVC
- Added FlowNetwork
- Created an UFF viewer executable
- Extended ImagePyramid to be able to create custom ImagePyramids
- Added SegmentationPyramidRenderer to render super high resolution segmentations stored as a pyramid.
- Added zip library and Qt Network to enable download of test data from software. Added executable downloadTestData which can be used to download test data. All test data is now put in C:/ProgramData/FAST/data/ and /home//FAST/data.
- Bux fixes etc ++
Release binaries and installers can be downloaded below, by pressing "assets". Please make sure you have the required dependencies installed on your system before using the release.
v3.1.0
Changelog
- Updated OpenVINO to version 2020.3
- Updated Clarius Listener API to version 7.1.0
- Updated TensorRT to version 7
- Updated Catch testing framework to version 2
- Updated Qt to version 5.14
- Visual Studio 2019 support
- Updated FAST to use C++ 17 (thus ubuntu 16.04 no longer supported (old gcc))
- Text based pipeline system improvements, see wiki for details
- runPipeline tool for running text pipelines
- HDF5 module with support for streaming data from the ultrasound file format (UFF)
- Bounding box detection neural network, rendering and data object
- Removed need for qt.conf
- Improved exceptions, printing them to console even on windows
- Terminal color eye candy
- Windows installer and debian package support for releases
- Bux fixes etc ++
Release binaries and installers can be downloaded below, by pressing "assets". Please make sure you have the required dependencies installed on your system before using the release.
v3.1.0-rc1
Changelog
- Updated OpenVINO to version 2020.1
- Updated TensorRT to version 7
- Updated Catch testing framework to version 2
- Updated Qt to version 5.14
- Visual Studio 2019 support
- Updated FAST to use C++ 17 (thus ubuntu 16.04 no longer supported (old gcc))
- Text based pipeline system improvements, see wiki for details
- HDF5 module with support for streaming data from the ultrasound file format (UFF)
- Removed need for qt.conf
- Improved exceptions, printing them to console even on windows
- Terminal color eye candy
- Windows installer and debian package support for releases
- Bux fixes etc ++
Release binaries can be downloaded below, by pressing "assets". Please make sure you have the required dependencies installed on your system before using the release.