Skip to content

v2.0

Latest
Compare
Choose a tag to compare
@Chrisqcwx Chrisqcwx released this 16 Jun 09:31
· 24 commits to main since this release
3b55dc3
refactor all (#59)

* refactor gans and base

* feat: add refactor for plg

* feat: add refactor for plg

* refactor: location of latents sampler

* fix: some error

* feat: add some comments

* fix some error & add some docs

* fix: detach

* fix: parallel error

* fix: parallel apply error

* refactor: image classifier model forward

* fix: hook error

* fix error of classifier and add train example

* fix: example error

* refactor: remove some unuse

* refactor: position of sampler

* refactor attacker

* feat: add preprocess datasets

* feat: ppa

* feat: add bido, vib, ls trainer

* fix: some error

* fix some error

* feat: add constraint

* feat: add brep optimization

* feat: add rlb & mirror white

* fix: out of memory

* minor refactor

* fix: warning when resize

* refactor: reformat

* fix: final evaluation

* fix: final selection

* fix: preprocess dataset error

* feat: dataset preprocess

* refactor: preprocess data

* fix: dataset preprocess

* fix: some error in examples

* feat: add feature loss & fix: some typo

* fix: description error for datasets

* fix: typo

* fix: typo and programma error

* fix: metric batch

* refactor lomma gmi

* fix: classifier training

* fix: train script for classifier

* feat: dataset readme

* fix: some error

* fix: dataset error

* fix: dataset

* add: original ppa code

* fix: remove unused code

* add: split files

* fix: some bugs

* remove datasets

* deleted:    datasets/READba/README.md
	deleted:    datasets/celeba/celeba_split.py
	deleted:    datasets/hdceleba/README.md
	deleted:    datasets/hdceleba/celeba_split.py
	renamed:    datasets/celeba/split_files/private_test.txt -> src/modelinversion/datasets/split_files/private_test.txt
	renamed:    datasets/celeba/split_files/private_train.txt -> src/modelinversion/datasets/split_files/private_train.txt
	renamed:    datasets/celME.md
	deleted:    datasets/celeeba/split_files/public.txt -> src/modelinversion/datasets/split_files/public.txt

* feat: add some deep inversion loss & fix some error

* fix: celeba train

* remove incv1

* fix:some bugs

* fix:log for classfier training

* iter in gan training

* minor modification

* modified:   examples/standard/attacks/gmi.py
	modified:   examples/standard/gan_training/gmi.py
	modified:   examples/standard/gan_training/plgmi.py
	modified:   src/modelinversion/sampler/base.py
	modified:   src/modelinversion/utils/hook.py

* fix: dataset process in classifier training

* modified:   examples/standard/attacks/gmi.py

* modified:   examples/standard/gan_training/gmi.py
	modified:   src/modelinversion/metrics/base.py

* fix: acc in trainer

* params in gmi

* fix:set center crop as true

* rename some classes

* fix: programma error in loss

* rename some classes

* remove BaseTargetModel

* feat: release some script & repr for loss

* fix: generate dataset by generator

* fix: reset params to <fill it>

* feat: raise exception in torchvision wrapper

* substitute Flatten with nn.Flatten

* substitute Flatten with nn.Flatten

* add ImageImageFolder declaration

* fix: typo

* fix: add abstract method to BaseImageGenerator

* fix: correct a spelling mistake

* fix: comments in scores

* refactor: dataset & add ffhq

* fix: ppa script error

* rename GanTrainer

* refactor gan

* feat: add some comments in utils

* add some comments in io.py

* add some comments in io.py

* add some comments in utils

* feat: generator dataset

* add acc var

* feat: add attack flow

* try fix: lokt gan

* feat: add register for models

* add register utils in __iit__

* feat: config mixin

* feat: register for model config

* feat: classifier wrapper config

* feat: classifier config load successfully

* feat: save config after train classifier

* feat: config for GAN

* feat: config in __init__

* save config in gan training

* refactor: location of config

* fix: vib config

* feat: comments on ConfigMixin

* fix: acc std

* fix: acc metric

* fix classifier

* feat: lokt generation

* fix: config error

* fix: save error

* refactor: remove some unuse

* fix: metric exception

* fix: plg gan

* fix: ked gan train & metric; feat: tl training

* fix: kedmi256 structure

* feat: ked train script

* refactor: rename some scripts

* fix: deep inversion first bn weight

* feat: add resnest

* feat: Add BaseOutput for Optimization

* fix: metric error

* feat: add face dist & fix std error

* fix: save metric resule

* feat: post metric

* refactor attacker

* fix: final eval

* fix: gather

* fix: final selection

* fix lomma gmi

* fix: final selection

* feat: add post evaluation

* feat: add top k score optimized in genetic optimization

* refactor: remove unuse package

* add vmi (#60)

* feat: add sampler

* modified:   examples/standard/attacks/vmi.py

* feat: add latent sampler

* feat: add optimization

* feat: add optimization

* feat: add optimization

* feat: add loss for vmi

* feat: add loss for vmi

* feat: prepare for attack

* feat: prepare for attack

* fix: remove num_range

* refactor: prepare single thread

* feat: add single thread vmi

* feat: add FlowConfig

* feat: add multithread training

* feat: add metrics

* feat: add metrics

* feat: add metrics

* feat: add metrics

* feat: add metrics

* feat: add transform

* fix: some minor bugs

---------

Co-authored-by: final-solution <[email protected]>

* fix: fid

* feat: c2f

* fix: c2f

* feat: defense example

* Supplement (#61)

* add: parameter management

* feat: extend the parameter management in the future

* feat: extend the parameter management in the future

* fix: new requirements

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* feat: README for examples/standard

* feat: README for examples/standard

---------

Co-authored-by: final-solution <[email protected]>
Co-authored-by: Yixiang Qiu <[email protected]>
Co-authored-by: Wenbo Yu <[email protected]>

---------

Co-authored-by: final-solution <[email protected]>
Co-authored-by: Yixiang Qiu <[email protected]>
Co-authored-by: Wenbo Yu <[email protected]>