search

ykdai / BasicPBC

Official Implementation of "Learning Inclusion Matching for Animation Paint Bucket Colorization"
Other
259 stars 23 forks source link

[Fea] all resolution and trappedball seg inference #9

Closed dienachtderwelt closed 7 months ago

dienachtderwelt commented 7 months ago

Update model inference by adding support for:

  1. input data of all resolutions
  2. unclosed line art images

Changes include:

  1. some up/downsampling within the model arch
  2. a new dataset that gives squared input data from all kinds of user resolutions
  3. some help functions/classes that perform trappedball segmentation and extract 'seg' for unclosed line art image
  4. more arguments in the inference script
  5. a simple test anime clip and more...

See code changes for details.