CNN Feature Extractor

• This code can be used to extract the CNN penultimate layer feature vectors from the state-of-the-art Convolutional neural network architectures which are trained on 1 million ImageNet images.

• The code shows the example of using RESNET-152 version 2. But it can be easily modified to use a different CNN architectures, provided in the models folder, which are originally developed by Tensorflow-slim authors.

  • ResNet 152 v2
  • ResNet 152 v1
  • VGG16