Change Log#

Radionets v0.4.0 (2025-08-06)#

API Changes#

Bug Fixes#

    • replaced variance with standard deviation [#155]

    • compute uncertainty histogram on source pixels only

    • check for existing sampling file before generating

    • proper error propagation after using a normalization [#164]

    • fix evaluate_msssim_sampled

    • fix call of evaluate_msssim_sampled [#171]

    • Fix legend handles missing in eval contour plot [#175]

Data Model Changes#

New Features#

    • added eval methods for sampled images [#155]

    • add zenodo.json [#161]

    • add new normalizing method which normalizes every image

    • add saving options for more evaluation methods [#165]

    • Added new block types for deep ResNets and UNets: BottleneckResBlock, Encoder, and Decoder

    • Introduced new submodules activation, archs, and blocks

      • Related classes (e.g., blocks, activation functions) and functions are moved from other submodules into these respective submodules

      • Improves readability and reusability

    • Added a logger that replaces bare prints throughout the code base

    • Added new diverging colormap radionets.PuOr that is used for plotting in radionets [#178]

    • Added docs with API references and user/dev guides [#187]

Maintenance#

  • use mamba in tests [#159]

    • add MANIFEST.in [#160]

    • clean up docs/changes [#163]

    • delete unused functions

    • delete unsued architectures

    • rename symmetry function [#166]

    • add radionets logo to README [#169]

    • fix comet_ml callback

    • update process_prediction with better if statement

    • change hardcoded values for sampling [#170]

    • set number of bins for histogram plotting [#171]

    • Update pyproject and python version support

    • Switch to hatchling build backend [#176]

    • Restructure dl_framework.architecture [#177]

    • Flattened module hierarchy

    • Refactord architecture into modular components, see new features

    • Moved LocallyConnected2d class to unc_archs

    • Refactored some callback submodule

      • Add error handling for cases where normalization attributes (self.learn.normalize.mode) may not be defined

      • Remove unnecessary calls to create_OrBu(), replace it with direct import of OrBu

    • Refactored plotting tools and grouped plotting functions into logical groups

    • CI: Replaced pip with uv for package installation and added codecov test analytics [#178]

Refactoring and Optimization#

    • add keyword for half of the image

    • distinguish between tensor and array in get_ifft

    • fix micromamba installation [#168]

Radionets 0.3.0 (2023-08-04)#

API Changes#

Bug Fixes#

  • Fixed loading of correct sampling file [#145]

  • Calculated normalization only on non-zero pixels

    • Fixed typo in rescaling operation [#149]

  • Fixed sampling for images displayed in real and imaginary part [#152]

New Features#

  • Enabled training and evaluation of half sized images (for 128 pixel images) [#140]

  • Added naming to save path, so that the files do not get overwritten as easily [#144]

  • Added normalization callback with two different techniques

    • Updated plotting routines for real/imag images

    • Updated evaluate_area and evaluate_ms_ssim for half images

    • Added evaluate_ms_ssim for sampled images [#146]

  • Add evaluation of intensity via peak flux and integrated flux comparison [#150]

  • Centered bin on 1 for histogram evaluation plots

    • Added color to legend [#151]

  • Added prettier labels and descriptions to plots [#152]

Maintenance#

  • Deleted unusable functions for new source types

  • Deleted unused hardcoded scaling [#140]

  • Added masked loss functions

  • Sorted bundles in simulations

  • Minor adjustments in plotting scripts [#141]

  • Consistent use of batch_size [#142]

  • Added the model name to predictions and sampling file

    • Deleted unnecessary pad_unsqueeze function

    • Added amp_phase keyword to sample_images

    • Fixed deprecation warning in sampling.py

    • Added image size to test_evaluation.py routines [#146]

  • Outsourced preprocessing steps in train_inspection.py [#148]

  • Removed unused norm_path from all instances [#153]

  • Deleted cropping

    • Updated colorbar label

    • Removed source_list argument [#154]

Refactoring and Optimization#

  • Optimized evaluation.utils.trunc_rvs with numba, providing functions compiled for cpu and parallel cpu computation. [#143]

Radionets 0.2.0 (2023-01-31)#

API Changes#

  • Train on half-sized iamges and applying symmetry afterward is a backward incompatible change

  • Models trained with early versions of radionets are not supported anymore [#140]

Bug Fixes#

  • Fixed sampling of test data set

  • Fixed same indices for plots [#140]

New Features#

  • Enabled training and evaluation of half sized images (for 128 pixel images) [#140]

Maintenance#

  • Deleted unusable functions for new source types

  • Deleted unused hardcoded scaling [#140]

Refactoring and Optimization#

Radionets 0.1.18 (2023-01-30)#

API Changes#

Bug Fixes#

New Features#

  • Added creation of uncertainty plots

  • Changed creation and saving/reading of predictions to dicts

    • Prediction dicts have 3 or 4 entries depending on uncertainty

  • Added scaled option to get_ifft

  • Created new dataset class for sampled images

  • Created option for sampling and saving the whole test dataset

  • Updated and wrote new tests [#129]

Maintenance#

  • Added and enabled towncrier in CI. [#130]

  • Published radionets on pypi [#134]

  • Updated README, used figures from the paper, minor text adjustments [#136]

Refactoring and Optimization#