PaDiM with wide_resnet50_2

[haobo827]

Only resnet18 in
model:
name: padim
backbone: resnet18
layers:
- layer1
- layer2
- layer3
normalization_method: min_max # options: [none, min_max, cdf]
threshold:
image_default: 3
pixel_default: 3
adaptive: true

If I want to use wide_resnet50_2 as net, how can I write the config.yaml?

[samet-akcay]

You need to set backbone: wide_resnet50_2

[haobo827]

You need to set backbone: wide_resnet50_2

Do other parameters in [model] need to be changed?

[samet-akcay]

It depends on what you want, but if you just want to choose wide_resnet50_2 as the backbone and extract the first three layers, you don't need to change anything.

[haobo827]

It depends on what you want, but if you just want to choose wide_resnet50_2 as the backbone and extract the first three layers, you don't need to change anything.

That's what I'm thinking at the moment, how nice of you

[isackf]

Hi, @samet-akcay, Is there a way to replace/change the extracted feature from layer1, layer2 and layer3 to another specific layer of the CNN backbone?

model:
name: padim
backbone: resnet18
layers:
- layer1
- layer2
- layer3

[djdameln]

Hi, yes this is possible. Just find the names of the layer(s) from which you want the model to extract the features, and specify them in the list. For example:

model:
  name: padim
  backbone: resnet18
  layers:
    - layer3.0
    - layer4

[isackf]

Got it, Thank you. Amazing work!