Gradient based adversarial attack (foolbox) for quantised deep learning models

[mohanrajroboticist]

• The salt and pepper noise attack from foolbox works for standard and quantised deep learning models.
• The gardient based attacks works only for standard deep learning models and doesnot work for quantised models.

The notebook has options to enable and disable quantisation. It also has options to change the type of attack.
https://colab.research.google.com/drive/182_vaLLvo8duOT3XZNp3gGCIJxdi-uA5?usp=sharing