Feature Extraction

[MehrdadHMT]

thank you for sharing this code,
i'm working on some image tagging problem based on CNN - RNN models. so,i used the .caffemodel file and classified the dataset but i need to extract image features that your code produces before the last InnerProduct layers and feed them into the RNN, but i don't know how.
can you help me please.
ps : I don't want to train the model again! just using the pre-trained model and instead of getting the labels, i need the feature vectors that (ip1 + sig1) layers are producing.