Prediction by deep learning
ImageDev exposes two new features for performing a prediction from a convolutional network model.
- OnnxPredictionFiltering2d computes a prediction on a two-dimensional image from an ONNX model and generates an image representing the prediction scores.
- OnnxPredictionSegmentation2d computes a prediction on a two-dimensional image from an ONNX model and applies a post processing to generate a label or a binary image.
Image denoising by deep learning prediction with a noise-to-noise model |
Membrane segmentation by deep learning prediction with a U-net model trained with the Avizo software |
ImageDev prediction tools use the Open Neural Network Exchange (ONNX) run time. ONNX is an interoperable framework enabling collaboration in the AI community. The ONNX framework provides tools for executing AI operations and a data model for representing convolutional neural networks.
Dental Panoramics
The DentalUnfolding3d new algorithm is the last step for extracting a dental panoramic from a CBCT volume.
It unfolds a 3D surface defined in an input mesh, to make it appear as a flat layer of the output 3D panoramic volume.
Initial dental CBCT and fitted mesh | Extracted panoramic |
This algorithm was the missing step for replacing the legacy DentalPanoramicExtractor class of ImageViz. It can now be totally implemented with ImageDev.
Gradient Commands
The new GradientVector2d and GradientMagnitude2d algorithms are now available. These algorithms are intended to replace the GradientOperator2d algorithm that is now deprecated.
The new GradientVector3d and GradientMagnitude3d algorithms are now available. These algorithms are intended to replace the GradientOperator3d algorithm that is now deprecated.
Measurement information
A new FieldInformation class has been added to get information about the dimension and unit of a given measurement. This class is used by to types objects:
- For any field of a global measurement data object, it is returned by the fieldNameInformation() method of the measurement object.
- For any Measurement object of an individual analysis, it is returned by the information() method .
Compatibility changes
The undocumented and unimplemented fieldNameInfo() methods of the Global Measurement data objects have been replaced by the fieldNameInformation() methods. These new methods are part of the official API and can be used.
New Supported Platforms
ImageDev is now available for:
- Visual Studio 2019 (VC 16) on Windows
- Ubuntu 20.04 (GCC 9) on Linux.
- Python 3.9 and 3.10 on Linux.
Deprecations
- The support of Ubuntu 18.04 will be stopped in ImageDev 2023.2.
- The Deblur2d algorithm is deprecated, it will be removed in ImageDev 2024.1. The Deblurring2d algorithm can be used instead.
- The GradientOperator2d algorithm is deprecated, it will be removed in ImageDev 2025.1. The GradientVector2d or GradientMagnitude2d algorithm can be used instead.
- The GradientOperator3d algorithm is deprecated, it will be removed in ImageDev 2025.1. The GradientVector3d or GradientMagnitude3d algorithm can be used instead.
Operating systems
ImageDev 2022.2 is available for
- Windows 64-bit: 2017 (VC 15) and 2019 (VC 16).
- Linux 64-bit: GCC 7 (Ubuntu 18.04) and GCC 9 (Ubuntu 20.04).
Future removals
In the next ImageDev versions, the following features will be removed.
ImageDev 2023.2
- End of Ubuntu 18.04 support on Linux.
ImageDev 2024.1
- Deblur2d: the Deblurring2d algorithm can be used instead.
ImageDev 2025.1
- GradientOperator2d: the GradientVector2d or GradientMagnitude2d algorithm can be used instead. The CANNY_DERICHE, SHEN_CASTAN, CANNY, and AMPLITUDE_AND_ORIENTATION modes have no equivalent in the new API. Please contact our support if it is an impediment for your developments.
- GradientOperator3d the GradientVector3d or GradientMagnitude3d algorithm can be used instead. The CANNY_DERICHE, and CANNY modes have no equivalent in the new API. Please contact our support if it is an impediment for your developments.