-
Notifications
You must be signed in to change notification settings - Fork 1
Bash Scripts
Below are links to other user created multiscale resolution/multiscale generation scripts, meant for style transfer.
Some modification may be required for scripts originally meant for neural-style if used with neural-style-pt, and vice versa:
For example the interpreters th and python or python3 will have to be swapped if you replace neural_style.lua with neural_style.py or vice versa. If you use neural-style-pt's pip package, then there is no need to use an interpreter, and you can just use neural-style. The -multigpu_strategy flag in neural-style was replaced with -multidevice_strategy in neural-style-pt. neural-style-pt does not have -normalize_gradients and it's TV loss is different than the original neural-style. neural-style-pt also does not have -proto_file, as PyTorch does not use prototxt files, and thus neural-style-pt attempts to get the model definition via the model file's name.
Creator: vic8760
-
VOLTA-X3: https://pastebin.com/jXVaYaYW, https://www.reddit.com/r/deepdream/comments/954h2w/voltax3_script_release_the_best_hq_neuralstyle/
-
SF-DELTA-G7: https://pastebin.com/0WsL8fnQ, https://www.reddit.com/r/deepdream/comments/c66bpa/sfdeltag7_script_release_improved_optimization_2x/
-
Uni8: https://pastebin.com/xmhRGRc8, https://www.reddit.com/r/deepdream/comments/8vuv5u/the_stanford_campus_uni8_script_prototype/
Creator: bododge