ONNX Neural Networks in Breeze
Import and run externally trained neural network models thanks to our support for the .ONNX file format that can be generated using software like Matlab and PyTorch.
Import and run externally trained neural network models thanks to our support for the .ONNX file format that can be generated using software like Matlab and PyTorch.
Run your own Python code for data processing steps in combination with the built in functions and methods in the software
Apply advanced image segmentation and object detection using a big selection of different methods that consider both the spectral and spatial information in your images.
Apply methods that use the spectral differences in your data to do image segmentation and detect objects, such as machine learning and chemometrics models as well as band method functions.
Using the spatial information related to the shape of objects, you can apply methods based on deep learning and neural networks. The Breeze software includes the pre-trained FastSAM neural network for shape-based analysis, which can be used directly without any need for additional training. Other models, such as YOLO and Faster R-CNN, must be created
Other segmentations based on pixel coordinates or manual selection of ROI are also available options. To make use of the data each pixel can provide, Breeze also includes segmentation options for dividing an object into subsections or selecting a desired amount of pixels from each object.
Streamline your data processing by building your own advanced workflows that enable fast and easy analysis of large data sets.
Add your data processing functions such as spectral pre-processing, image segmentation, models and algorithms, outputs and statistics. The user-friendly menu with a large selection of different functions makes this easy and quick.
A visual representation of your data processing workflow makes it easy and intuitive to add, remove, or update the different data processing steps. At Prediktera, we call this an Analysis Tree – A tree structure where you can build your own workflow by adding functions to the nodes. The data processing workflow can run both
Quickly run your workflow one or a few images to test your analysis results, or apply it on hundreds of images with one click. View the results from the analysis in an interactive table with numbers and images.