myria3d

Getting Started

  • Install Myria3D on Linux
    • Setting up a virtual environment
    • Install source as a package
    • Troubleshooting
  • Install Myria3D on WSL2 with CUDA support
    • Setting up WSL2
    • Installing Anaconda
    • Installing Myria3D
    • Install cuda in WSL
    • Troubleshooting
  • Preparing data for training
    • Peprocessing functions
    • Preparing the dataset
    • Getting started quickly with a toy dataset
  • Performing inference on new data
    • Run inference from source
    • Run inference from sources
    • Run inference from within a docker image
    • Additional options for prediction

Guides

  • How to train new models
    • Setup
    • Quick run
    • Training
    • Testing the model
    • Inference
  • Developer’s guide
    • Code versionning
    • Tests
    • Continuous Integration (CI)
    • Continuous Delivery (CD)

Background

  • KNN-Interpolation to merge multiple predictions [TODO]
  • General design of the package
    • Model should be fast, performant, and practical
    • Subsampling is important to improve point cloud structure
    • Speed is of the essence
    • Evaluation is key to select the right approach

Package Reference

  • Scripts
    • run
    • myria3d.train
    • myria3d.predict
  • Default configuration
  • myria3d.pctl
    • myria3d.pctl.datamodule.hdf5
    • myria3d.pctl.dataset.hdf5
    • myria3d.pctl.dataset.iterable
    • myria3d.pctl.dataset.toy_dataset
    • myria3d.pctl.dataset.utils
    • myria3d.pctl.dataloader.dataloader
    • myria3d.pctl.points_pre_transform.lidar_hd
    • myria3d.pctl.transforms.compose
    • myria3d.pctl.transforms.transforms
  • myria3d.models
    • Model
    • Interpolation
  • myria3d.models.modules
    • (Pytorch-Geometric) RandLA-Net
  • myria3d.callbacks
    • Submodules
    • myria3d.callbacks.comet_callbacks
    • myria3d.callbacks.finetuning_callbacks
    • myria3d.callbacks.logging_callbacks
    • Module contents
  • myria3d.utils
    • myria3d.utils.utils
myria3d
  • »
  • Search

© Copyright 2024, Institut National de l'Information Géographique et Forestière.

Built with Sphinx using a theme provided by Read the Docs.