search

vineeths96 / Spoken-Keyword-Spotting

In this repository, we explore using a hybrid system consisting of a Convolutional Neural Network and a Support Vector Machine for Keyword Spotting task.
MIT License
90 stars 18 forks source link

question #11

Closed SamanaBS closed 3 years ago

SamanaBS commented 3 years ago

why is the OCSVM taking input from the penultimate dense layer of CNN ?

vineeths96 commented 3 years ago

We use the CNN model as a feature extractor. Since we have trained a model to perform word classification, we expect it to learn rich latent representations. Hence, we pick the output of a dense layer as our feature for OCSVM.

Infact, you can use any layer's output as your feature. You should probably try with different ones. But typically, the deeper the better

SamanaBS commented 3 years ago

okay thank