samet-akcay / ganomaly Goto Github PK

View Code? Open in Web Editor NEW

840.0 840.0 211.0 75 KB

GANomaly: Semi-Supervised Anomaly Detection via Adversarial Training

License: MIT License

Python 98.57% Shell 1.43%

anomaly-detection gan pytorch semi-supervised-learning

ganomaly's Introduction

ganomaly's People

Contributors

Stargazers

Watchers

Forkers

baiyancheng20 bemoregt qingsong99 tiffen zhouyonglong wzmsltw gearchen chaooma xushjie1987 yifan-guo-cwru yuehchang lss616263 ahong286 aevilorz jianyuheng tartaruszen hdyen speakingtom jjy0710 kutim djtimy bianjiang1234567 aarthipriyar liyaoshun justmyfantasy lwpyh yoheikikuta jaykimbravekjh jiweitian evilangels zhouzhou12 phuccuongngo99 cooleel catherinezll95 boxoq-rex zhiqiang-zhu yaowang-bjtu ailib ptsoucas wangerniuniu anormaly-detection lurenlym yuanyuanzijin gdjmck snowisland1 satchelwu yui010206 elementai zzuiekongning oziris johnson13 jameshyunkim stwrd hangtianhengxing nullees uzeroj andrew05200 gerardmaggiolino matak07 stjordanis ruyueshuo youtang1993 djxuelei babyjia arifmudi kevinpiger draganraja wzg16 vedfu facelsss vishalbelsare jianqiangc maweifei fudanyuan2019 kmvanhespen waveletpytorch seanyan604 leinok darcyqin pkusnail ericphilbaz jjkke88 pkq1688 gheyret xuanli98 warden3344 zxpjustin kmtgifu eggpan95 yiran-hao pancakeawesome neelbhowmik lyra7417 chalot lxmwust tramp00 15529523003 silencesengoku yixiyixi5 wuzujiong

ganomaly's Issues

Is Discriminator loss correct??

Hi, I'm trying to understand the mechanism of GANomaly.

I read model.py and I thought Discriminator Loss was incorrect because it should maximize Ladv or minimize BCE.

Is Discriminator loss correct??

Something wrong for CIFAR experiments

HI! I run your code with pytorch0.3 + python3.5. I tried to run the demo in experiments/run_cifar.sh. But I met some index error. Error shows like follows:
Traceback (most recent call last):
File "train.py", line 47, in
main()
File "train.py", line 44, in main
model.train()
File "/media/yangbiao/disk1/ganomaly-master/lib/model.py", line 307, in train
self.train_epoch()
File "/media/yangbiao/disk1/ganomaly-master/lib/model.py", line 272, in train_epoch
for data in tqdm(self.dataloader['train'], leave=False, total=len(self.dataloader['train'])):
File "/usr/local/lib/python3.5/dist-packages/tqdm/_tqdm.py", line 719, in iter
for obj in iterable:
File "/usr/local/lib/python3.5/dist-packages/torch/utils/data/dataloader.py", line 281, in next
return self._process_next_batch(batch)
File "/usr/local/lib/python3.5/dist-packages/torch/utils/data/dataloader.py", line 301, in _process_next_batch
raise batch.exc_type(batch.exc_msg)
IndexError: Traceback (most recent call last):
File "/usr/local/lib/python3.5/dist-packages/torch/utils/data/dataloader.py", line 55, in _worker_loop
samples = collate_fn([dataset[i] for i in batch_indices])
File "/usr/local/lib/python3.5/dist-packages/torch/utils/data/dataloader.py", line 55, in
samples = collate_fn([dataset[i] for i in batch_indices])
File "/usr/local/lib/python3.5/dist-packages/torchvision/datasets/cifar.py", line 83, in getitem
img, target = self.train_data[index], self.train_labels[index]
IndexError: index 49369 is out of bounds for axis 0 with size 45000
Do you know why? I will appreciate your aid.

some questions about Custom Dataset

Hello,I am reading your code now,and I want to ask you that if the Custom Dataset have labels like minst label 1,2,3,4,....
I am trying to do somthing that classify the normal class.
Hope for your relpy.

Is the number of epochs task-sensitive?

I am a newbie and interested in this repository. I am curious that if we can set a large number of epochs to attain better AUC?

Discriminator Loss Specification

Amazing work! Just a comment, in the paper its not very clear if to train the discriminator you where going to:

Maximize Ladv for a (real, fake) pair
Minimize the cross_entropy loss for D(x) = 1 and D(G(x)) = 0

In the repo I see you use cross entropy but it would be good to clarify it in the paper, since you don't mention this its hard to understand the adversarial aspect of the training method.

What are dimensionalities of images in UBD and FFOB catasets?

I see that many people ask you to release datasets, but, unfortunately you cannot do that. That's fine.

I want to ask you how big the images are? Are we talking about 500x500 pixels? Or even 1024x1024? I haven't found this information in the paper.

Thank you.

error: expected input[64, 1, 32, 32] to have 3 channels, but got 1 channels instead

hi, when I switched to Linux with Nvdia GPU, I met a new problem. when I use command like this:

python3 train.py --dataset mnist --niter 20 --anomaly_class 5
python3 train.py --dataset cifar10 --niter 20 --anomaly_class cat

I tried on mnist and cifar 10, the result is the same like this:

root@z-Precision-Tower-3620:/home/z/桌面/git/ganomaly#` python3 train.py --dataset mnist --niter 20 --anomaly_class 5
/usr/local/lib/python3.6/dist-packages/sklearn/externals/joblib/externals/cloudpickle/cloudpickle.py:47: DeprecationWarning: the imp module is deprecated in favour of importlib; see the module's documentation for alternative uses
  import imp
/usr/local/lib/python3.6/dist-packages/torchvision/transforms/transforms.py:188: UserWarning: The use of the transforms.Scale transform is deprecated, please use transforms.Resize instead.
  "please use transforms.Resize instead.")
>> Training model Ganomaly.
Traceback (most recent call last):                                              
  File "train.py", line 43, in <module>
    model.train()
  File "/home/z/桌面/git/ganomaly/lib/model.py", line 279, in train
    self.train_epoch()
  File "/home/z/桌面/git/ganomaly/lib/model.py", line 249, in train_epoch
    self.optimize()
  File "/home/z/桌面/git/ganomaly/lib/model.py", line 181, in optimize
    self.update_netd()
  File "/home/z/桌面/git/ganomaly/lib/model.py", line 134, in update_netd
    self.out_d_real, self.feat_real = self.netd(self.input)
  File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 477, in __call__
    result = self.forward(*input, **kwargs)
  File "/home/z/桌面/git/ganomaly/lib/networks.py", line 153, in forward
    features = self.features(x)
  File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 477, in __call__
    result = self.forward(*input, **kwargs)
  File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/container.py", line 91, in forward
    input = module(input)
  File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 477, in __call__
    result = self.forward(*input, **kwargs)
  File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/conv.py", line 301, in forward
    self.padding, self.dilation, self.groups)
RuntimeError: Given groups=1, weight of size [64, 3, 4, 4], expected input[64, 1, 32, 32] to have 3 channels, but got 1 channels instead

Can you help me to fix this? Thx!

There are some mistakes in GANomaly paper

In paper 3.2 Model Training, the Encoder Loss part :
1)In last sentence which is write "'Lcon is formally defined as Lenc=...." ,It's a mistake i think,the 'Lcon' should be corrected to 'Lenc'.
2) In second sentence "encoder loss Lcon to minimize the..." should corrected 'Lcon' to Lenc too.

Prediction

Could you please include the instruction for doing the prediction?

ValueError: multiclass format is not supported

Hi，I tried your new update code 。but I got this error

Traceback (most recent call last):
File "train.py", line 43, in
train()
File "train.py", line 40, in train
model.train()
File "/home/dl/VSST/dm/ganomaly-master/lib/model.py", line 168, in train
res = self.test()
File "/home/dl/VSST/dm/ganomaly-master/lib/model.py", line 259, in test
auc = evaluate(self.gt_labels, self.an_scores, metric=self.opt.metric)
File "/home/dl/VSST/dm/ganomaly-master/lib/evaluate.py", line 24, in evaluate
return roc(labels, scores)
File "/home/dl/VSST/dm/ganomaly-master/lib/evaluate.py", line 46, in roc
fpr, tpr, _ = roc_curve(labels, scores)
File "/home/dl/anaconda3/envs/pytorch0.4/lib/python3.7/site-packages/sklearn/metrics/ranking.py", line 622, in roc_curve
y_true, y_score, pos_label=pos_label, sample_weight=sample_weight)
File "/home/dl/anaconda3/envs/pytorch0.4/lib/python3.7/site-packages/sklearn/metrics/ranking.py", line 396, in _binary_clf_curve
raise ValueError("{0} format is not supported".format(y_type))
ValueError: multiclass format is not supported

Issue in displaying

Hi,

Thanks for sharing this code. When I tried to include the display option in train.py. I get the following error

raise ConnectionError(e, request=request)
requests.exceptions.ConnectionError: HTTPConnectionPool(host='localhost', port=8097): Max retries exceeded with url: /events (Caused by NewConnectionError('<urllib3.connection.HTTPConnection object at 0x7f77db544bd0>: Failed to establish a new connection: [Errno 111] Connection refused'))

Use of visdom server is not being helpful. Any suggestions to solve this will be greatly appreciated.
Also, Please share if you added new resources for the prediction/ testing.

Thanks in advance!

How to test mnist and cifar10 datasets?

I don't see the test.py file in the folder. Do I need to write a test.py file to test the mnist and cifar10 data set? Or what parameters can I modify to test?
anyone can help me? thank you !

where is the test code?

After training, how could we test a single image to classify whether it's fake or real without using test() in model.py?

I tried you code in mnist dataset, but I got the following error:

can you upload the code of skip ganomaly

I have read your paper of skip ganomaly, your results are so good. Can you release you code so that I can learn from that? Besides, your paper use the different loss compared with the loss in ganomaly.

test() function is correct?

I tested cifar10(abnormal_class: bird) with the latest commit code which solved the loss problem.

As a result, a maximum of 0.579(AUC) was obtained, which was different from the GANomaly-bird score of 0.510 on paper(skip-ganomaly).

So while checking the source code, I found that self.netg.eval() is missing from the test() function in lib/model.py

So I added self.netg.eval() to 197 line and retested to get a 0.410 AUC.
It was quite different from the results of the paper.

I think it's right to test by adding eval(). What do you think?
If adding eval() is right, what do you think about the evaluation results of the paper?

I met this error. When train time/

Hi, @samet-akcay

I met this error.

File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 169, in add_module
raise KeyError("module name can't contain "."")
KeyError: 'module name can't contain "."'

What's wrong with me?

Thanks in advance.

from @bemoregt.

AttributeError: can't set attribute

Hi, there are something wrong when i try to run your codes:
File "E:/software/installation/Anaconda3/envs/CleverHans_py3.5/Lib/site-packages/ganomaly/train.py", line 35, in
dataloader = load_data(opt)
File "E:\software\installation\Anaconda3\envs\CleverHans_py3.5\Lib\site-packages\ganomaly\lib\data.py", line 99, in load_data
abn_cls_idx=opt.anomaly_class
AttributeError: can't set attribute

I don't know the ture reason, can you help me?
Thank you.

Broken pipe for demo sh

Hi~
I'm a student and impressed of your paper, then I decided to give it a try. But I'm facing some unknown problem when running the demo sh "run_mnist.sh" and "run_cifar.sh". (I moved these two sh file to the folder containing train.py.)

results for cifar10:

`Manual Seed: 0
Running CIFAR. Anomaly Class: plane
C:\Users\foo\AppData\Local\Continuum\anaconda3\lib\site-packages\torchvision\transforms\transforms.py:188: UserWarning: The use of the transforms.Scale transform is deprecated, please use transforms.Resize instead.
"please use transforms.Resize instead.")
Files already downloaded and verified
Files already downloaded and verified

Training model Ganomaly.
0%| | 0/703 [00:00<?, ?it/s]C:\Users\foo\AppData\Local\Continuum\anaconda3\lib\site-packages\torchvision\transforms\transforms.py:188: UserWarning: The use of the transforms.Scale transform is deprecated, please use transforms.Resize instead.
"please use transforms.Resize instead.")
Files already downloaded and verified
Files already downloaded and verified
Training model Ganomaly.
0%| | 0/703 [00:00<?, ?it/s]Traceback (most recent call last):
File "", line 1, in
File "C:\Users\foo\AppData\Local\Continuum\anaconda3\lib\multiprocessing\spawn.py", line 105, in spawn_main
exitcode = _main(fd)
File "C:\Users\foo\AppData\Local\Continuum\anaconda3\lib\multiprocessing\spawn.py", line 114, in _main
prepare(preparation_data)
File "C:\Users\foo\AppData\Local\Continuum\anaconda3\lib\multiprocessing\spawn.py", line 225, in prepare
_fixup_main_from_path(data['init_main_from_path'])
File "C:\Users\foo\AppData\Local\Continuum\anaconda3\lib\multiprocessing\spawn.py", line 277, in _fixup_main_from_path
run_name="mp_main")
File "C:\Users\foo\AppData\Local\Continuum\anaconda3\lib\runpy.py", line 263, in run_path
pkg_name=pkg_name, script_name=fname)
File "C:\Users\foo\AppData\Local\Continuum\anaconda3\lib\runpy.py", line 96, in _run_module_code
mod_name, mod_spec, pkg_name, script_name)
File "C:\Users\foo\AppData\Local\Continuum\anaconda3\lib\runpy.py", line 85, in _run_code
exec(code, run_globals)
File "C:\Users\foo\Desktop\git\ganomaly\train.py", line 43, in
model.train()
File "C:\Users\foo\Desktop\git\ganomaly\lib\model.py", line 279, in train
self.train_epoch()
File "C:\Users\foo\Desktop\git\ganomaly\lib\model.py", line 244, in train_epoch
for data in tqdm(self.dataloader['train'], leave=False, total=len(self.dataloader['train'])):
File "C:\Users\foo\AppData\Local\Continuum\anaconda3\lib\site-packages\tqdm_tqdm.py", line 937, in iter
for obj in iterable:
File "C:\Users\foo\AppData\Local\Continuum\anaconda3\lib\site-packages\torch\utils\data\dataloader.py", line 501, in iter
return _DataLoaderIter(self)
File "C:\Users\foo\AppData\Local\Continuum\anaconda3\lib\site-packages\torch\utils\data\dataloader.py", line 289, in init
w.start()
File "C:\Users\foo\AppData\Local\Continuum\anaconda3\lib\multiprocessing\process.py", line 105, in start
self._popen = self._Popen(self)
File "C:\Users\foo\AppData\Local\Continuum\anaconda3\lib\multiprocessing\context.py", line 223, in _Popen
return _default_context.get_context().Process._Popen(process_obj)
File "C:\Users\foo\AppData\Local\Continuum\anaconda3\lib\multiprocessing\context.py", line 322, in _Popen
return Popen(process_obj)
File "C:\Users\foo\AppData\Local\Continuum\anaconda3\lib\multiprocessing\popen_spawn_win32.py", line 33, in init
prep_data = spawn.get_preparation_data(process_obj._name)
File "C:\Users\foo\AppData\Local\Continuum\anaconda3\lib\multiprocessing\spawn.py", line 143, in get_preparation_data
_check_not_importing_main()
File "C:\Users\foo\AppData\Local\Continuum\anaconda3\lib\multiprocessing\spawn.py", line 136, in _check_not_importing_main
is not going to be frozen to produce an executable.''')
RuntimeError:
An attempt has been made to start a new process before the
current process has finished its bootstrapping phase.

    This probably means that you are not using fork to start your
    child processes and you have forgotten to use the proper idiom
    in the main module:

        if __name__ == '__main__':
            freeze_support()
            ...

    The "freeze_support()" line can be omitted if the program
    is not going to be frozen to produce an executable.

Traceback (most recent call last):
File "train.py", line 43, in
model.train()
File "C:\Users\foo\Desktop\git\ganomaly\lib\model.py", line 279, in train
self.train_epoch()
File "C:\Users\foo\Desktop\git\ganomaly\lib\model.py", line 244, in train_epoch
for data in tqdm(self.dataloader['train'], leave=False, total=len(self.dataloader['train'])):
File "C:\Users\foo\AppData\Local\Continuum\anaconda3\lib\site-packages\tqdm_tqdm.py", line 937, in iter
for obj in iterable:
File "C:\Users\foo\AppData\Local\Continuum\anaconda3\lib\site-packages\torch\utils\data\dataloader.py", line 501, in iter
return _DataLoaderIter(self)
File "C:\Users\foo\AppData\Local\Continuum\anaconda3\lib\site-packages\torch\utils\data\dataloader.py", line 289, in init
w.start()
File "C:\Users\foo\AppData\Local\Continuum\anaconda3\lib\multiprocessing\process.py", line 105, in start
self._popen = self._Popen(self)
File "C:\Users\foo\AppData\Local\Continuum\anaconda3\lib\multiprocessing\context.py", line 223, in _Popen
return _default_context.get_context().Process._Popen(process_obj)
File "C:\Users\foo\AppData\Local\Continuum\anaconda3\lib\multiprocessing\context.py", line 322, in _Popen
return Popen(process_obj)
File "C:\Users\foo\AppData\Local\Continuum\anaconda3\lib\multiprocessing\popen_spawn_win32.py", line 65, in init
reduction.dump(process_obj, to_child)
File "C:\Users\foo\AppData\Local\Continuum\anaconda3\lib\multiprocessing\reduction.py", line 60, in dump
ForkingPickler(file, protocol).dump(obj)
BrokenPipeError: [Errno 32] Broken pipe
`results for mnist:`
Manual Seed: 0
Running mnist_0
C:\Users\foo\AppData\Local\Continuum\anaconda3\lib\site-packages\torchvision\transforms\transforms.py:188: UserWarning: The use of the transforms.Scale transform is deprecated, please use transforms.Resize instead.
"please use transforms.Resize instead.")
Exception in user code:

Traceback (most recent call last):
File "C:\Users\foo\AppData\Local\Continuum\anaconda3\lib\site-packages\urllib3\connection.py", line 141, in _new_conn
(self.host, self.port), self.timeout, **extra_kw)
File "C:\Users\foo\AppData\Local\Continuum\anaconda3\lib\site-packages\urllib3\util\connection.py", line 83, in create_connection
raise err
File "C:\Users\foo\AppData\Local\Continuum\anaconda3\lib\site-packages\urllib3\util\connection.py", line 73, in create_connection
sock.connect(sa)
ConnectionRefusedError: [WinError 10061] ▒▒▒▒Ŀ▒▒▒▒▒▒▒▒▒▒▒ܾ▒▒▒▒޷▒▒▒▒ӡ▒