camelop / RL-fooling-framework

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