Skip to content

Releases: raghakot/keras-vis

Bug fixes and perf improvements in 0.4

06 Jul 04:56
Compare
Choose a tag to compare
Pre-release
  • Improved performance of guided backprop with subsequent calls.
  • various minor bug/doc fixes.

guided backprop, API improvements.

05 Jul 08:58
Compare
Choose a tag to compare
Pre-release
  • Introducing grad_modifiers, backprop_modifiers, input_modifiers.
  • Implemented guided and rectified backprop.
  • API changes. Deprecated visualize_regression variant and removed RegressionTarget loss,
  • Added visualize_xxx_with_losses variant for flexibility.
  • Moved all examples to notebooks.
  • Introduced backend abstraction for backend specific code.
  • Modularized visualization code.

Bug fixes

29 Jun 04:36
Compare
Choose a tag to compare
Bug fixes Pre-release
Pre-release

Fixes various minor issues in 0.3.2

Introducing regression visualizations.

28 Jun 21:13
Compare
Choose a tag to compare
Pre-release
  • Implemented regression variant for all visualizations.
  • Refactored common code for various visualizations for advanced use cases of specifying custom loss.
  • Added RegressionTarget experimental loss function.
  • Added an example use-case for visualizing a toy self-driving car.
  • Added the ability to specify input range for activation visualizations
  • Decoupled heatmap overlay in attention visualizations as not all inputs are images.

Various bug fixes

19 Jun 03:35
Compare
Choose a tag to compare
Various bug fixes Pre-release
Pre-release
0.3.1

Update PyPI

N-dim support

17 May 19:47
Compare
Choose a tag to compare
N-dim support Pre-release
Pre-release
  • Added support for N-dim attention, and activation maximization
  • Huge perf improvements
  • Removed a ton of unwanted dependencies.

Added attention maps

25 Dec 01:00
Compare
Choose a tag to compare
Added attention maps Pre-release
Pre-release

API changes, added attention related visualizations and examples.

Initial Release

24 Dec 01:47
Compare
Choose a tag to compare
Initial Release Pre-release
Pre-release

Supports

  • Conv/Dense layer guided backprop
  • Saliency maps
  • grad-CAM