trushkin-ilya / face-anti-spoofer Goto Github PK

View Code? Open in Web Editor NEW

48.0 48.0 10.0 135.58 MB

Deep Learning based Face Anti-Spoofing

License: Apache License 2.0

Python 100.00%

anti-spoofing computer-vision deep-learning pytorch

face-anti-spoofer's People

Contributors

Stargazers

Watchers

Forkers

brahimbellahcen bholdmanny anthonyyuan lwppwl wuxiaolianggit chaoso hhxylcy3 codesharecw agusk gziren

face-anti-spoofer's Issues

Add --num_workers parameter to data loading

Train script doesn't save the model

Traceback (most recent call last):
  File "submit.py", line 43, in <module>
    model.load_state_dict(torch.load(getattr(args,f'model{protocol}_path'), map_location=device.type))
  File "/home/itrushkin/.virtualenvs/antispoofing/lib/python3.7/site-packages/torch/nn/modules/module.py", line 803, in load_state_dict
    state_dict = state_dict.copy()
AttributeError: 'function' object has no attribute 'copy'

Baseline model link broken

Can you share code for baseline models that are used for training the checkpoints, the sym-link in the repo seems to be broken.

Crash in SplittedDataLoader

File "train.py", line 44, in <module>
    dataloader = utils.SplittedDataLoader(dataset, args.train_batch_size, args.val_batch_size, args.num_workers)
  File "/home/agruzdev/git/face-anti-spoofer/utils.py", line 7, in init
    val_q = len(dataset) // sum(split) * split[1]
TypeError: 'int' object is not iterable

Allow to load model accross device types

Traceback (most recent call last):
  File "submit.py", line 43, in <module>
    model.load_state_dict(torch.load(getattr(args,f'model{protocol}_path')))
  File "/home/itrushkin/.virtualenvs/antispoofing/lib/python3.7/site-packages/torch/serialization.py", line 529, in load
    return _legacy_load(opened_file, map_location, pickle_module, **pickle_load_args)
  File "/home/itrushkin/.virtualenvs/antispoofing/lib/python3.7/site-packages/torch/serialization.py", line 702, in _legacy_load
    result = unpickler.load()
  File "/home/itrushkin/.virtualenvs/antispoofing/lib/python3.7/site-packages/torch/serialization.py", line 665, in persistent_load
    deserialized_objects[root_key] = restore_location(obj, location)
  File "/home/itrushkin/.virtualenvs/antispoofing/lib/python3.7/site-packages/torch/serialization.py", line 156, in default_restore_location
    result = fn(storage, location)
  File "/home/itrushkin/.virtualenvs/antispoofing/lib/python3.7/site-packages/torch/serialization.py", line 132, in _cuda_deserialize
    device = validate_cuda_device(location)
  File "/home/itrushkin/.virtualenvs/antispoofing/lib/python3.7/site-packages/torch/serialization.py", line 116, in validate_cuda_device
    raise RuntimeError('Attempting to deserialize object on a CUDA '
RuntimeError: Attempting to deserialize object on a CUDA device but torch.cuda.is_available() is False. If you are running on a CPU-only machine, please use torch.load with map_location=torch.device('cpu') to map your storages to the CPU.

Cannot extract named parameter from args while submitting

Traceback (most recent call last):
  File "submit.py", line 43, in <module>
    model.load_state_dict(torch.load(args[f'model{protocol}_path']))
TypeError: 'Namespace' object is not subscriptable

Trained model for testing purpose

Hi Author,

Any Chance you guys are releasing trained-model as well.

Or the CASIA-SURF dataset itself. ?

Changing --batch_size doesn't change training process

Batch size parameter isn't passed to get_loader method:

face-anti-spoofer/src/train.py

Line 39 in a1daa43

train_loader, val_loader = get_loader(dataset)

No module named tqdm

Traceback (most recent call last):
  File "submit.py", line 7, in <module>
    from tqdm import tqdm
ModuleNotFoundError: No module named 'tqdm'

Fix evaluation interval

face-anti-spoofer/src/train.py

Line 65 in 928352a

if epoch % args.eval_every == 1:

Division remainder should be equal to zero.

Model not work well with intel realsense d435

I downloaded weights and set up running with intel realsense d435, but get poor performance. I know casia-surf dataset is captured by intel real sense sr300 . So, it may not work with d435. If you have solution or idea , please help me

New Pillow version breaks torchvision

pytorch/vision#1712