Feature extraction blocks

vision
1

How can I see which blocks have what kind of layers? like mobilenetv3 has 5 downsampling blocks (5 stages) if one want to use as a backbone. But how can I know what is in stage 1, stage 2, stage 3, stage 4, and stage 5?

I have extracted the features of it using the below code:

from torchvision import models

mobilenet = models.mobilenet_v3_large(pretrained=True)
features = mobilenet.features

Can anyone guide? @ptrblck

2

You can print the corresponding layers in your script, check the implementation in its source code, or could also try to visualize the model in e.g. TensorBoard.
I usually start by printing the interesting modules and then dig into the code to see how these layers are used.