deeptarget.py

import numpy as np

def deeptarget(image, f, grads, overshoot=0.02, max_iter=50,target=None):

    """
       :param image: Image of size HxWx3
       :param f: feedforward function (input: images, output: values of activation BEFORE softmax).
       :param grads: gradient functions with respect to input (as many gradients as classes).
       :param overshoot: used as a termination criterion to prevent vanishing updates (default = 0.02).
       :param max_iter: maximum number of iterations for deepfool (default = 10)
       :return: minimal perturbation that fools the classifier, number of iterations that it required, new estimated_label and perturbed image
    """
    if target==None:
        raise ValueError("Target is not enough")

    f_image = np.array(f(image)).flatten()
    I = (np.array(f_image)).flatten().argsort()[::-1]

    I = I[0:2]
    label = I[0]
    I[1]=target

    input_shape = image.shape
    pert_image = image

    f_i = np.array(f(pert_image)).flatten()
    k_i = int(np.argmax(f_i))
    o_i=k_i

    w = np.zeros(input_shape)
    r_tot = np.zeros(input_shape)

    loop_i = 0

	# we can't use k_i!=target. DeepTarget is limited L2 norm .therefore generate huge perturbation.
    while k_i == label and loop_i < max_iter:

        pert = np.inf
        gradients = np.asarray(grads(pert_image,I))


            # set new w_k and new f_k
        w = gradients[1, :, :, :, :] - gradients[0, :, :, :, :]
        f_t = f_i[I[1]] - f_i[I[0]]
        pert = abs(f_t)/np.linalg.norm(w.flatten())

        # compute r_i and r_tot
        r_i =  pert * w / np.linalg.norm(w)
        r_tot = r_tot + r_i

        # compute new perturbed image
        pert_image = image + (1+overshoot)*r_tot
        loop_i += 1

        # compute new label
        f_i = np.array(f(pert_image)).flatten()
        k_i = int(np.argmax(f_i))

    r_tot = (1+overshoot)*r_tot
    #print(o_i," --> ",k_i)

    return r_tot, loop_i, k_i, pert_image