arivis Vision4D 3.5 Release Notes August 2021
Summary of important changes in arivis Vision4D 3.5 which was released 2nd Aug 2021.
Top new features:
- New Object Styles to define object's color, transparency, material, and representation by tag
- Analysis Home screen with better access to sample pipelines
- Machine Learning Object Classification and performance improvements in all ML tools (incl. GPU/CUDA support)
- reworked Compartments operation supporting parent-child relationship with subjects within a certain distance and fast feature definition per level of hierarchy
- analysis pipeline operations implemented as Python script
- arivis VisionHub integration
General:
- "What's new" intro screen
- arivis VisionHub integration (up and download of pipelines, images and objects from/to VisionHub)
- significantly faster SIS save (structural modifications)
- Fix: incorrect scalebar length at certain pixel size ratios
- Windows 7 is no longer supported
Import & Data Handling:
- better support for import of HCA well plate datasets:
- Zeiss CZI well plate datasets
- GE XDCE datasets
- Yokogawa MLF and MMD datasets
- support for import of ND2 float data
- improved Nikon ND2 import using the latest ND2Read SDK
- better support for importing Leica LEI time series
- Olympus import using the latest IDAL library
- improved edit support for tiles in manual mode of the Tile Sorter
- much faster handling of a very large number of files to import
- Fix: rare issue where old metadata prevented SIS from opening
- Fix: Tile Sorter showing projection will update thumbnail after changing z-shift of a tile
4D Viewer:
- New glass material style for object visualization
- improved and simplified lightings settings
- unified clipping planes front/back orientation
- 360° video export in 4K resolution
- Fix: end marker for 3D polyline objects
- Fix: prevent wrong positioned objects created with the Drop Objects tool in MIP view mode
Analysis:
- Improved Import Document Objects operation
- Analysis Home Screen provides a well-organized starting point for your analyses
- reworked input/output handling with support for multiple outputs from one operation
- Tracking provides secondary output for tracked segments and optional output for non tracked
- Seeded Region Growing provides optional output for the seeds
- reworked and redesigned Compartments operation
- Compartments operation supports creating parent-child relationships with subjects within a certain distance
- Compartments operation may classify objects as inside, intersecting, or proximal to the reference
- Compartments operation provide optional outputs for different relation subgroups
- The Compartments operation now supports the easy creation of Custom Features (e.g. count of children or mean intensity) for contained children
- Combine Outputs operation supports more than two inputs
- Machine Learning Object Classification analysis operation with different classes as separate outputs
- Visualize the disk location for temporary analysis files in the status bar
- Warning if the free disk space might be exceeded on running an analysis pipeline
- changed: No auto-select of a different input channel if the configured one is missing (warning for existing pipelines not fitting the data)
- A new analysis pipeline operations to use a Python script to
-
- run a voxel operation or to
- create or to
- Modify segments
Performance:
- improved: speed up segment creation for 2D data by avoiding surface precalculations
- much increased performance when creating or modifying complex segment structures
Batch analysis:
- allow to switch to batch analysis with the pipeline just configured
- remember last batch analysis output folder
- improved batch analysis configuration guidance
- changed: use all image sets by default in batch analysis when adding new files
- Fix: avoid potential problems working with a large number of objects (>10 million)
- Fix: prevent unintentional jumping of the view during execution of the analysis pipeline step by step
- Fix: preview for the Region Growing operation
- Fix: potential problem with inconsistent error messages
- Fix: avoid potential hang of the Membrane Detection operation
- Fix: bug in Image Math which could cause strange pattern
- Fix: potential crash with the ML segmenter
- Fix: prevent missing entries in the Export Object Features operation (edge cases)
- Fix: problem with saving the pipeline if some special Custom Compute Features are involved
Machine Learning:
- New Machine Learning Object Classification Trainer
- Run Machine Learning Object Classification to classify visible objects (no Analysis Pipeline needed)
- ML Trainer (Image) allows to paint into a class other than the one displayed
- ML Trainer (Image) - warning about 3D-features preview inaccuracy
- ML Segmentation provides classes as separate outputs in the analysis pipeline
- Machine Learning image operations benefit from CUDA hardware acceleration (if available)
- Much improved Machine Learning (Image) performance
Objects:
- New Object Styles allow to define the visual style of objects more easily including color, opacity, material and representation (surface, centroid or approximation)
- Setting of objects styles by tag with the new Object Styles Dialog
- Support for individual transparencies, colors and display type for single objects
- Option to ignore scope offsets while copying objects
- Tags are shown per image set
- Fix: rare problem with the magic wand mode of the Manual Track tool
- Fix: show color by time or by segment mode indication in the object table
- Fix: prevent unexpected disappearance of feature columns
Scripting:
- Integrated debugger for the Python script editor
- SegmentBuilder script support
- Dedicated scripting log
- Preference to define script log verbosity
- Scripting of object styles
- Clean up of script API
- More detailed script API documentation
- Improved Script editor help pane UI
- Fix: potential problem in handling pixel size