Source code for perceptron.benchmarks.additive_noise

# Copyright 2019 Baidu Inc.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

"""Metric that tests models against differnt types of additive noise."""

from abc import abstractmethod
from collections import Iterable
from tqdm import tqdm
import numpy as np
from .base import Metric
from .base import call_decorator
from perceptron.utils.rngs import nprng


[docs]class AdditiveNoiseMetric(Metric): """Base class for metric that tests models against additive noise."""
[docs] @call_decorator def __call__(self, adv, annotation=None, unpack=True, abort_early=True, epsilons=10000): """Adds uniform or Gaussian noise to the image, gradually increasing the standard deviation until the image is misclassified. Parameters ---------- adv : `numpy.ndarray` or :class:`Adversarial` The original, unperturbed input as a `numpy.ndarray` or an :class:`Adversarial` instance. annotation : int The reference label of the original input. Must be passed if `a` is a `numpy.ndarray`, must not be passed if `a` is an :class:`Adversarial` instance. unpack : bool If true, returns the adversarial input, otherwise returns the Adversarial object. abort_early : bool If true, returns when got first adversarial, otherwise returns when all the iterations are finished. epsilons : int or Iterable[float] Either Iterable of standard deviations of the Gaussian blur or number of standard deviations between 0 and 1 that should be tried. """ a = adv del adv del annotation del unpack image = a.original_image bounds = a.bounds() min_, max_ = bounds if not isinstance(epsilons, Iterable): epsilons = np.linspace(0, 1, num=epsilons + 1)[1:] for epsilon in tqdm(epsilons): noise = self._sample_noise(epsilon, image, bounds) perturbed = image + epsilon * noise perturbed = np.clip(perturbed, min_, max_) _, is_adversarial = a.predictions(perturbed) if is_adversarial and abort_early: return
@abstractmethod def _sample_noise(self): raise NotImplementedError
[docs]class AdditiveUniformNoiseMetric(AdditiveNoiseMetric): """Metric that tests models against uniform noise.""" def _sample_noise(self, epsilon, image, bounds): min_, max_ = bounds w = epsilon * (max_ - min_) noise = nprng.uniform(-w, w, size=image.shape) noise = noise.astype(image.dtype) return noise
[docs]class AdditiveGaussianNoiseMetric(AdditiveNoiseMetric): """Metric that tests models against Gaussian noise.""" def _sample_noise(self, epsilon, image, bounds): min_, max_ = bounds std = epsilon / np.sqrt(3) * (max_ - min_) noise = nprng.normal(scale=std, size=image.shape) noise = noise.astype(image.dtype) return noise