Skip to content

Releases: dan64/vs-deoldify

Version v5.0.3

18 Feb 14:47
Compare
Choose a tag to compare

Maintenance release.

Added:

  • new API function: HAVC_merge
    this a utility function with the implementation of HAVC merge methods
  • new VapourSynth sample scripts in HAVC User Guide using new APIs

Version v5.0.2

16 Feb 19:24
Compare
Choose a tag to compare

Maintenance release.

Added:

  • new parameter value "Custom" to HAVC_bw_tune
  • new VapourSynth sample scripts in HAVC User Guide

Release v5.0.1

14 Feb 17:08
Compare
Choose a tag to compare

This is maintenance release.

Since DeepRemaster requires access to the future colored frames, its use is restricted to the following methods:

  • 3 = external RF same as video
  • 4 = external RF different from video
  • 5 = external ClipRef same as video
  • 6 = external ClipRef different from video

Version v5.0.0

12 Feb 10:26
Compare
Choose a tag to compare

This release introduce significant improvements and new functionalities.

The main changes introduced are:

  • Added new exemplar-based model: DeepRemaster
  • Added new functionality to use external colored movies as reference image source
  • Added new functionality to restore old colored movies with DeepRemaster
  • Added new APIs supporting the new functionalities
  • Updated the User Guide with the description of the new functionalities and more examples
  • Added the folder samples with some clips and reference images that can be used to test the filter.

Version v4.6.8

19 Jan 19:46
Compare
Choose a tag to compare

Small maintenance release.

  • The scene change detection algorithm has been improved to properly handle low thresholds levels.
  • The chapter 4.0.2 of HAVC user guide has been update to provide examples using low thresholds levels.

Version v4.6.7

18 Jan 17:46
Compare
Choose a tag to compare

The aim of this release is to remove the excessive red components that are often used to color the frames.
Unfortunately all the picture color models are affected by this problem.
The changes in this release will affect some Presets.

The main changes are the following:

  • Preset Color Map added/renamed some color mapping as displayed in the following table:

Preset_ColorMap

  • Preset Ref merge: added new weight as described below
    the new Ref Merge allowed values are:
    0 = No RF merge (reference frames can be produced with any frequency)
    1 = RF-Merge VeryLow (reference frames are merged with weight=0.3)
    2 = RF-Merge Low (reference frames are merged with weight=0.4)
    3 = RF-Merge Med (reference frames are merged with weight=0.5)
    4 = RF-Merge High (reference frames are merged with weight=0.6)
    5 = RF-Merge VeryHigh (reference frames are merged with weight=0.7)

  • HAVC User Guide

    • Improved chapter 3.5 with some useful example
    • Added chapter 4.0.2 Best settings to remove colors shifting towards red

Version v4.6.5

12 Jan 10:37
Compare
Choose a tag to compare

Respect to the previous version were applied the following changes:

Function: HAVC_main()

  1. removed from the list of allowed parameters in VideoTune: 'DeOldify', 'DDColor'
  2. added new parameter ColorModel, with the following allowed values:
    'Video+Artistic' (default)
    'Video+ModelScope'
    'Video+Siggraph17'
    'Video+ECCV16'
    'DeOldify(Video)'
    'DDColor(Artistic)'
    'DDColor(ModelScope)'
    'Zhang(Siggraph17)'
    'Zhang(ECCV16)'

This change was necessary to give more flexibility to select the color models (which were increased in the last version) using the Presets in HAVC_main.

Version v4.6.1

05 Jan 13:27
Compare
Choose a tag to compare

Maintenance release: some small bug fixes and code refactoring.

Version v4.6.0

03 Jan 11:12
Compare
Choose a tag to compare

Added 2 new color models as an alternative to DDcolor: Real-Time User-Guided Image Colorization with Learned Deep Priors (Zhang, 2017) and Colorful Image Colorization (Zhang, 2016), these 2 models are named in Hybrid as: siggraph17, eccv16, they have in common the same settings and tweaks of DDcolor.

On first use the following pytorch models will be downloaded:

In vs-deoldify the torch hub directory (used to store the pytorch models) is in: Hybrid\64bit\Vapoursynth\Lib\site-packages\vsdeoldify\models

Release v4.5.2

29 Dec 15:58
Compare
Choose a tag to compare
  • Added utility function HAVC_export_reference_frames
  • Fixed typos in functions description