change feature extractors-替换特征提取网络

[xuzhengzhenghit]

When I tried to replace Padim feature extraction from resnet18 with mobilenetv3_large_100, I encountered a bug：

D:\appacations\Anaconda\envs\my_anomalib\lib\site-packages\torchmetrics\utilities\prints.py:36: UserWarning: Metric `PrecisionRecallCurve` will save all targets and predictions in buffer. For large datasets this may lead to large memory footprint.
  warnings.warn(*args, **kwargs) 2024-03-20 16:08:26,019 - anomalib.models.components.feature_extractors.timm - WARNING - FeatureExtractor is deprecated. Use TimmFeatureExtractor instead. Both FeatureExtractor and TimmFeatureExtractor will be removed in a future release.
D:\appacations\Anaconda\envs\my_anomalib\lib\site-packages\anomalib\models\components\feature_extractors\timm.py:82: UserWarning: Layer blocks.1.1 not found in model mobilenetv3_large_100
  warnings.warn(f"Layer {i} not found in model {self.backbone}") 2024-03-20 16:08:26,234 - timm.models.helpers - INFO - Loading pretrained weights from url (https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/mobilenetv3_large_100_ra-f55367f5.pth) Traceback (most recent call last):
  File "tools/train.py", line 79, in <module>
    train(args)
  File "tools/train.py", line 58, in train
    model = get_model(config)
  File "D:\appacations\Anaconda\envs\my_anomalib\lib\site-packages\anomalib\models\__init__.py", line 132, in get_model
    model = getattr(module, f"{convert_snake_to_pascal_case(config.model.name)}Lightning")(config)
  File "D:\appacations\Anaconda\envs\my_anomalib\lib\site-packages\anomalib\models\padim\lightning_model.py", line 123, in __init__
    super().__init__(
  File "D:\appacations\Anaconda\envs\my_anomalib\lib\site-packages\anomalib\models\padim\lightning_model.py", line 49, in __init__
ractors\utils.py", line 26, in <dictcomp>
    layer: {"num_features": dryrun_features[layer].shape[1], "resolution": dryrun_features[layer].shape[2:]}
KeyError: 'blocks.2.2'

my config:

model:
  name: padim
  backbone: mobilenetv3_large_100 #resnet18 wide_resnet50_2
  pre_trained: true
  layers:
    - blocks.1.1
    - blocks.2.2
  normalization_method: min_max # options: [none, min_max, cdf]

This is my reference website：

https://anomalib.readthedocs.io/en/latest/markdown/guides/how_to/models/feature_extractors.html#available-backbones-and-layers

将Paiim特征提取网络换了，出现报错，操作步骤参考了官方的指导文档
救命家人们

[abc-125]

Hello, in the feature extractors docs you can see the warning, "Some models might not support every backbone". I am afraid it can be the problem here, Padim simply doesn't support MobileNet backbones.

[blaz-r]

That is quite unusual, as the docs have this specific net and I think last time it worked for me. I'll check this and get back to you.