Skip to content

Latest commit

 

History

History
75 lines (71 loc) · 3.78 KB

CHANGELOG.md

File metadata and controls

75 lines (71 loc) · 3.78 KB
Raw

Change log

NNDT 0.0.3 (13 Nov 2022)

  • Space model of the first version is purged
  • Updated tutorials and two major examples
  • nndt.datasets for download of toy datasets
  • Short imports
  • More methods support tensors of any shape
  • New data generators for out-of-the-box solvers: DataGenForShapeRegression, DataGenForSegmentation
  • Added new color iterations for .plot(), and new color scheme for .print()
  • New methods
    • Surface data (+2): .surface_xyz(), .surface_rgba()
    • Space model modification (+2): split_node_kfold, split_node_namelist
    • Visualization (+2): save_sdt_as_obj, save_3D_slices
  • Code quality
    • Pre-commit actions and the black codestyle
    • CI pipeline become faster for new PR
    • stable main, and dev branch for new features
    • 78% of code coverage in CodeCov
    • 27% of docstring coverage
  • Known issues
    • Wiki page is outdated
    • .plot() crush Google Colaboratory
    • Some bugs with load of .ir1 files
    • Some test is disabled because of CI pipeline restrictions
    • A- grade, 33 issues in CodeFactor

NNDT 0.0.2 (16 Oct 2022)

  • The partial becomes complete; the crooked, straight; the empty, full; the worn out, new.
  • Now we have 7 contributors. Thank you all for new docs, features, issues and CI/CD pipeline!
  • The space model (space) is pandas for 3D objects. It is completely redesigned and is now available as space2. The previous version of space is still here, but will be removed soon
  • Visualization of tree and 3D objects via .plot() and .print()
  • New short and easy import for space2
  • 30 methods are available in the second version of the space model
    • Available as functions (12)
      • Space model loaders (6): load_from_path, load_from_file_lists, load_only_one_file, load_txt, load_sdt, load_mesh_obj
      • Space model serialization (2): to_json, save_space_to_file
      • Space model deserialization (2): read_space_from_file, from_json
      • Primitive generation (1): add_sphere
      • Space model modification (1): split_node_test_train
    • Available as methods in the space model tree (18)
      • General (3): .plot, .print, .unload_from_memory
      • Sampling (4): .sampling_eachN_from_mesh, .sampling_grid, .sampling_grid_with_noise, .sampling_uniform
      • Surface data (6): .surface_ind2rgba, .surface_ind2xyz, .surface_xyz2ind, .surface_xyz2localsdt, .surface_xyz2rgba, .surface_xyz2sdt
      • Transformations (4): .transform_sdt_ns2ps, .transform_sdt_ps2ns, .transform_xyz_ns2ps, .transform_xyz_ps2ns
      • Save (1): .save_mesh
  • New modules for Haiku: MLP with Lipschitz regularization (LipMLP, LipLinear)
  • Functions for iteration over N-simplex with barycentric coordinates in math_core.py
  • Serialization and deserialization of the space2 model
  • Conversion of signed distance tensor (SDT) to signed distance function (SDF) using MLP training (still uses the previous version of space model)
  • One real demo for a research paper (still uses the previous version of space model)
    • Shape interpolation for building a computational anatomy atlas
  • Draft documentation for some methods
  • Code quality
    • CI pipeline for new pull requests
    • 74% of code coverage in CodeCov
    • A grade, 10 issues in CodeFactor
  • Known issues
    • Wiki page is outdated
    • Lack of documentation
    • Bugs with SDT requests and cube marching
    • Some test is disabled because of CI pipeline restrictions

NNDT 0.0.1 (16 Sep 2022)

  • A journey of a thousand miles begins with a single step.
  • The first release of the code with all key features.
  • Wiki page for maintainers and contributors.
  • Notebook tutorials:
    • Space models
    • Trainable tasks
  • Demos:
    • Mesh segmentation
    • Shape regression
    • Eikonal in 3D model