s_len = X_0.shape[0]
results.loc[idx].X_adv_err = torch.zeros(len(noise_rel), s_len)
results.loc[idx].X_ref_err = torch.zeros(len(noise_rel), s_len)
for s in range(s_len):
print("Sample: {}/{}".format(s + 1, s_len))
X_0_s = X_0[s:s + 1, ...].repeat(it_init,
*((X_0.ndim - 1) * (1, )))
Y_0_s = Y_0[s:s + 1, ...].repeat(it_init,
*((Y_0.ndim - 1) * (1, )))
X_adv_err_cur, X_ref_err_cur = grid_attack(
method,
noise_rel,
X_0_s,
Y_0_s,
store_data=False,
keep_init=keep_init,
err_measure=err_measure,
)
results.loc[idx].X_adv_err[:, s] = X_adv_err_cur.max(dim=1)[0]
results.loc[idx].X_ref_err[:, s] = X_ref_err_cur.mean(dim=1)
# save results
for idx in results.index:
results_save.loc[idx] = results.loc[idx]
os.makedirs(save_path, exist_ok=True)
results_save.to_pickle(save_results)
# ----- plotting -----
*((X_0.ndim - 1) * (1, )))
Y_0_s = Y_0[s:s + 1, ...].repeat(it_init,
*((Y_0.ndim - 1) * (1, )))
(
X_adv_err_cur,
X_ref_err_cur,
X_adv_cur,
X_ref_cur,
_,
_,
) = grid_attack(
method,
noise_rel,
X_0_s,
Y_0_s,
store_data=True,
keep_init=keep_init,
err_measure=err_measure,
)
(
results.loc[idx].X_adv_err[:, s],
idx_max_adv_err,
) = X_adv_err_cur.max(dim=1)
results.loc[idx].X_ref_err[:, s] = X_ref_err_cur.mean(dim=1)
for idx_noise in range(len(noise_rel)):
idx_max = idx_max_adv_err[idx_noise]
results.loc[idx].X_adv_psnr[idx_noise, s] = psnr(
rotate_real(X_adv_cur[idx_noise, ...])[idx_max, 0:1, ...],
