danielroich / PTI

Official Implementation for "Pivotal Tuning for Latent-based editing of Real Images" (ACM TOG 2022) https://arxiv.org/abs/2106.05744
MIT License
897 stars 112 forks source link

use_locality_regularization = False? #31

Closed lastapple closed 2 years ago

lastapple commented 2 years ago

In your code "./configs/hyperparameters.py", the "use_locality_regularization" is False? The space_regulizer_loss is not calculated ?

danielroich commented 2 years ago

For this specific configuration, space_regulizer_loss is not calculated. You can activate it by changing the value to True. If you plan to re-use the generator then I recommend changing the value, it will keep the W space intact.

The value is False by default because it enables faster inversion for experiments and one time usages

lastapple commented 2 years ago

Thank you for your quick reply!As the loss is your contribution, I used to wonder why this parameter was set to false

rmokady commented 2 years ago

lastapple

This is also useful when multiple images inverted simultaneously, but most users, especially non-expert, usually invert one image at a time.