Get 2048 Length Feature Vectors of Inception Model TorchVision 0.7.0

vision
1

I am attempting to train an LSTM for predicting video frames. The frames will be represented as feature vectors pulled from a the outputs. I am trying to get the 2048 size features right before classification, but am struggling to do so. When I use IntermediateLayerGetter, I recieve this error, when trying to run the new model.

RuntimeError: Expected 4-dimensional input for 4-dimensional weight [192, 768, 1, 1], but got 2-dimensional input of size [1, 1000] instead

Any hints as to what I am doing wrong or how to get these features?

1 Like
2

Could you post an executable code snippet, which shows your current approach?
You should be able to get the intermediate activation using forward hooks as described here.