The open source mesh and visualization tool Gmsh provides a general method for the post-processing of high-order finite element fields. However, the lack of visualization tools able to handle a large number of degrees of freedom (dof) has been a major bottleneck for the analysis of high-order finite element solutions generated by massively parallel simulations. Any contributions you make are greatly appreciated.Recently high-order finite element methods initially introduced in the research community such as discontinuous Galerkin (DG) and flux reconstruction (FR) have gained considerable attention in industry, thanks to their high-accuracy on unstructured meshes, their efficiency and scalability. └── Rough ideas # Initial idea/experiment with Programmable FilterĬontributions are what make the open source community such an amazing place to learn, inspire, and create. └── requirements.txt # package requirements │ └── img-segment # pre-trained models architecture for img segmentation │ ├── img-classify # pre-trained models architecture for img classification │ ├── fluid-segment # pre-trained models architecture for fluid segmentation │ ├── fluid-classify # pre-trained model architecture for fluid classification ├── pre-trained-models # all pre-trained models
PARAVIEW PLUGINS CODE
├── plugin-test # source code for test plugins (ground truth) │ ├── img_segment_plugin.py # plugin code for image segmentation │ ├── img_classifier_plugin.py # plugin code for image classification │ ├── fluid_segment_plugin.py # plugin code for fluid segmentation │ ├── fluid_classifier_plugin.py # plugin for fluid classification ├── plugin-src # main directory with source code of plugins Original Velocity Magnitude View in Paraview:Ībove data Segmented by Plugin in Paraview: The "Surface view" results of applying Fluid Segmentation Plugin for velocity data of a given timestep in Paraview are below. Tensorboard -logdir model-training/runs/ from the outer directory.
To check the graph of training & validation loss, run. Python train_fluid_segment.py for fluid segmentation, and the models Run python train_fluid_classifier.py for fluid classifier, and. Re-train the model for fluid classification and segmentation, follow the instructions below. The models for image classification and segmentation were downloaded from PyTorch, rest were trained by author. The pre-trained models are already provided in Releases. The test plugins are to see the ground truth results of computed fluid classification and segmentation, and are used to compare with the ML based classification & segmentation. The test plugins in this repository are named as follows: The main plugins in this repository are named as follows: Type the name of the filter (named below). After loading your input source, Go to Filters -> Search. Open you source input file in Paraview by File -> Open. Load all plugins required, eg, plugin-src/fluid_classifier_plugin.py. Find the location of the code of plugins in your disk (here, inside plugin-src/ directory). Go to Tools -> Manage Plugins -> Load New. In the conda environment you built, open Paraview. Instructions to use filter plugin in Paraview Some packages are not available in conda so those packages are listed in pip-requirements and are to be Make sure to turn on PARAVIEW_USE_PYTHON option while building it to enableįor installation of other packages and libraries:3.7.4Ĭonda create -name python=3.7.4 -file requirements.txt The plugins have been developed with Paraview 5.8.1 built against Python 3.8.5.įollow instructions to build Paraview here. Done By -ĭrishti Maharjan - Instructions for installation This repository contains four plugins for the use cases: image segmentation, image classification, fluid classification & fluid segmentation respectively. These plugins extend the filters in Paraview to include algorithms based on data-driven transformations. The filters builtin Paraview, till the time of writing, are based on deterministic algorithms to modify or extract slices of input data. These plugins transform the input data by feeding it into the model and then visualize the model’s output. Pre-trained means that machine learning models have been already trained on some form of dataset and carry the resultant weights and biases. To couple the interface of Paraview, a visualization application, and PyTorch, a machine learning framework, I have developed plugins in Paraview that allow users to load pre-trained models of their choice. Directory Structure inside 'thesis-code'. Instructions to use filter plugin in Paraview. Plugins for data-driven filters in Paraview Table of Contents
0 kommentar(er)
