kero Changelog

Version 0.6.3

NeuralNetwork.py. initiate_neural_network(). Add UniformRandom mode.

Version 0.6.2

NeuralNetwork.py: included NetworkUpdater class that performs model training.
utils.py: more functions are included.

Version 0.6.0

Included kero.utils.
Included NeuralNetwork.py in kero.multib.

Version 0.5.3

Included MANIFEST.in for sdist in Pypi upload.

Version 0.5.2

kero/multib/nDnet.py. Fixed bugs on

  1. SpecialGallery(). quiver3D_lattice3D, grid_shape=None assumes [N,1] shape.
  2. tuple_to_grid(). When Z is None, NoneType has no tolist(). problem will no longer occur.

Version 0.5.1

kero/multib/nDnet.py

quiver3D_lattice3D bug on squashed 3D plot fixed.

Version 0.5.0

kero/multib added.

Version 0.4.4

kero/ImageProcessing/photobBox.py

clone_to_annotate_faster() additional arguments 1. discard_edge and 2. edge_offset. Also fix bug on labeling object outside image or zero-width object.

Version 0.4.3

kero/ImageProcessing/photobBox.py
GreyScaleWorkshop:

  1. clone_to_annote () and Clone_to_annote_faster(). No longer need to set skip_ground_truth to False in order to use check_missing_mode = True
  2. scaled_box+positions() return None if there is no white patch RGB(255,255,255) found.
  3. multiple_scaled_box_positions() auto adjusts ground truth image rescaling if previous rescaling causes

tag_rename_clone: Added logging feature showing the change of names.