/RL-fooling-framework

Primary LanguagePython

RL-fooling-framework

preparation

download Mnist dataset and unzip all so the folder looks like:

  • dataset

    • original
      • t10k-images.idx3-ubyte
      • t10k-labels.idx1-ubyte
      • train-images.idx3-ubyte
      • train-labels.idx1-ubyte

  • pretrained model

    • model
      • mnist
        • checkpoint
          • LeNet
      • cifar10
        • checkpoint
          • MobileNet
          • VGG19

run experiment

  • To test the fooling, try python src/run.py 0
  • To convert a trajectory, try python src/trajectory_convert.py $Trajectory_location