• easily use google colab to run Homework_2.ipynb
• if your want to use your local environment to run this code, make sure you have tensorflow 2.x (that is because we use google colab to save the model, resulting in model files which are not compatible with tensorflow1.x)

1. down this repo
2. Download the validation and test datasets from here and put the .h5 datafile into /data:

├── data 
    └── cl
        └── valid.h5 // this is clean validation data used to design the defense
        └── test.h5  // this is clean test data used to evaluate the BadNet
    └── bd
        └── bd_valid.h5 // this is sunglasses poisoned validation data
        └── bd_test.h5  // this is sunglasses poisoned test data

3. open this project in google colab, and run Homework_2.ipynb, after run all code, you will see the repaired networks under /goodnets, goodnet_2, goodnet_4, goodnet_10 denote the repaired networks for X={2%, 4%, 10%}, respectively.

I have provided the code of the evaluation at the end of Homework_2.ipynb, you can check the evaluation results or run it with google colab