CIFAR models

.github/workflows/test-notebooks.yml

@@ -3,7 +3,7 @@ on: push
 jobs:
   build-n-deploy:
     name: Test 🔧
-    runs-on: ubuntu-18.04
+    runs-on: ubuntu-22.04
     steps:
     - name: Checkout code 🛎️
       uses: actions/checkout@v2.3.1

surrogates_overview/scripts/cifar_label_map.py

@@ -0,0 +1,166 @@
+"""
+CIFAR 10 & 100 Labels Map
+=========================
+
+This module provides a map from a class id to label.
+Two maps are available:
+
+* ``CIFAR10_LABEL_MAP`` -- maps class ids to labels for CIFAR10; and
+* ``CIFAR100_LABEL_MAP`` -- maps class ids to labels for CIFAR100.
+
+The data set files needed to regenerate the label maps are available at
+<https://www.cs.toronto.edu/~kriz/cifar.html>.
+
+See <https://github.com/fat-forensics/resources/tree/master/surrogates_overview>
+for more details.
+"""
+# Author: Kacper Sokol <k.sokol@bristol.ac.uk>
+# License: new BSD
+
+
+import pickle
+
+
+def _load_cifar10_labels(data_folder):
+    """Generates the label map for CIFAR10."""
+    with open(f'{data_folder}/cifar-10-batches-py/batches.meta', 'rb') as fo:
+        cf10meta = pickle.load(fo, encoding='bytes')
+
+    cf10labels = {i: j.decode()
+                  for i, j in enumerate(cf10meta.get(b'label_names'))}
+
+    return cf10labels
+
+
+def _load_cifar100_labels(data_folder, fine_labels=True):
+    """Generates the label map for CIFAR100."""
+    with open(f'{data_folder}/cifar-100-python/meta', 'rb') as fo:
+        cf100meta = pickle.load(fo, encoding='bytes')
+
+    if fine_labels:
+        cf100_labels_type = b'fine_label_names'
+    else:
+        cf100_labels_type = b'coarse_label_names'
+
+    cf100labels = {i: j.decode()
+                  for i, j in enumerate(cf100meta.get(cf100_labels_type))}
+
+    return cf100labels
+
+
+CIFAR10_LABEL_MAP = {
+    0: 'airplane',
+    1: 'automobile',
+    2: 'bird',
+    3: 'cat',
+    4: 'deer',
+    5: 'dog',
+    6: 'frog',
+    7: 'horse',
+    8: 'ship',
+    9: 'truck'
+}
+
+
+CIFAR100_LABEL_MAP = {
+    0: 'apple',
+    1: 'aquarium_fish',
+    2: 'baby',
+    3: 'bear',
+    4: 'beaver',
+    5: 'bed',
+    6: 'bee',
+    7: 'beetle',
+    8: 'bicycle',
+    9: 'bottle',
+    10: 'bowl',
+    11: 'boy',
+    12: 'bridge',
+    13: 'bus',
+    14: 'butterfly',
+    15: 'camel',
+    16: 'can',
+    17: 'castle',
+    18: 'caterpillar',
+    19: 'cattle',
+    20: 'chair',
+    21: 'chimpanzee',
+    22: 'clock',
+    23: 'cloud',
+    24: 'cockroach',
+    25: 'couch',
+    26: 'crab',
+    27: 'crocodile',
+    28: 'cup',
+    29: 'dinosaur',
+    30: 'dolphin',
+    31: 'elephant',
+    32: 'flatfish',
+    33: 'forest',
+    34: 'fox',
+    35: 'girl',
+    36: 'hamster',
+    37: 'house',
+    38: 'kangaroo',
+    39: 'keyboard',
+    40: 'lamp',
+    41: 'lawn_mower',
+    42: 'leopard',
+    43: 'lion',
+    44: 'lizard',
+    45: 'lobster',
+    46: 'man',
+    47: 'maple_tree',
+    48: 'motorcycle',
+    49: 'mountain',
+    50: 'mouse',
+    51: 'mushroom',
+    52: 'oak_tree',
+    53: 'orange',
+    54: 'orchid',
+    55: 'otter',
+    56: 'palm_tree',
+    57: 'pear',
+    58: 'pickup_truck',
+    59: 'pine_tree',
+    60: 'plain',
+    61: 'plate',
+    62: 'poppy',
+    63: 'porcupine',
+    64: 'possum',
+    65: 'rabbit',
+    66: 'raccoon',
+    67: 'ray',
+    68: 'road',
+    69: 'rocket',
+    70: 'rose',
+    71: 'sea',
+    72: 'seal',
+    73: 'shark',
+    74: 'shrew',
+    75: 'skunk',
+    76: 'skyscraper',
+    77: 'snail',
+    78: 'snake',
+    79: 'spider',
+    80: 'squirrel',
+    81: 'streetcar',
+    82: 'sunflower',
+    83: 'sweet_pepper',
+    84: 'table',
+    85: 'tank',
+    86: 'telephone',
+    87: 'television',
+    88: 'tiger',
+    89: 'tractor',
+    90: 'train',
+    91: 'trout',
+    92: 'tulip',
+    93: 'turtle',
+    94: 'wardrobe',
+    95: 'whale',
+    96: 'willow_tree',
+    97: 'wolf',
+    98: 'woman',
+    99: 'worm'
+}

surrogates_overview/scripts/image_classifier.py

@@ -2,15 +2,19 @@
 Image Classifier
 ================
 
-This module implements an image classifier based on PyTorch.
-Inception v3 and AlexNet are availabel.
+This module implements image classifiers based on PyTorch.
+Inception v3 and AlexNet are available for ImageNet;
+ResNet56 is available for CIFAR10; and
+RepVGG (a2) is available for CIFAR100.
+
 See <https://github.com/fat-forensics/resources/tree/master/surrogates_overview>
 for more details.
 """
 # Author: Kacper Sokol <k.sokol@bristol.ac.uk>
 # License: new BSD
 
 from scripts.imagenet_label_map import IMAGENET_LABEL_MAP
+from scripts.cifar_label_map import CIFAR10_LABEL_MAP, CIFAR100_LABEL_MAP
 
 import numpy as np
 
@@ -33,6 +37,26 @@ def _get_preprocess_transform():
     return transf
 
 
+def _get_preprocess_transform_cifar10():
+    # https://github.com/chenyaofo/pytorch-cifar-models/issues/4
+    # https://github.com/chenyaofo/image-classification-codebase/blob/master/conf/cifar10.conf
+    normalize = transforms.Normalize(
+        mean=[0.4914, 0.4822, 0.4465], std=[0.2023, 0.1994, 0.2010])
+    transf = transforms.Compose([transforms.ToTensor(), normalize])
+
+    return transf
+
+
+def _get_preprocess_transform_cifar100():
+    # https://github.com/chenyaofo/pytorch-cifar-models/issues/4
+    # https://github.com/chenyaofo/image-classification-codebase/blob/master/conf/cifar100.conf
+    normalize = transforms.Normalize(
+        mean=[0.5070, 0.4865, 0.4409], std=[0.2673, 0.2564, 0.2761])
+    transf = transforms.Compose([transforms.ToTensor(), normalize])
+
+    return transf
+
+
 class ImageClassifier(object):
     """Image classifier based on PyTorch."""
 
@@ -128,3 +152,77 @@ def proba2tuple(self, Y, labels_no=5):
                 tuples_.append((lab, Y[idx, cls], cls))
             tuples.append(tuples_)
         return tuples
+
+
+class ImageNetClassifier(ImageClassifier):
+    """ImageNet classifiers -- Inception v3 & AlexNet -- based on PyTorch."""
+
+
+class Cifar10Classifier(ImageClassifier):
+    """CIFAR10 classifiers -- ResNet56 -- based on PyTorch."""
+
+    def __init__(self, use_gpu=False):
+        """Initialises the image classifier."""
+        # Get class labels
+        self.class_idx = CIFAR10_LABEL_MAP
+
+        # Get the model
+        # https://github.com/huyvnphan/PyTorch_CIFAR10
+        clf = torch.hub.load(
+            'chenyaofo/pytorch-cifar-models',
+            'cifar10_resnet56',
+            pretrained=True)
+
+        if use_gpu:
+            if CUDA_AVAILABLE:
+                clf = clf.to(DEVICE)
+                # clf.cuda()
+                predict_proba = self._predict_proba_gpu
+            else:
+                logger.warning('GPU was requested but it is not available. '
+                               'Using CPU instead.')
+                predict_proba = self._predict_proba_cpu
+        else:
+            predict_proba = self._predict_proba_cpu
+        self.predict_proba = predict_proba
+
+        self.clf = clf
+        self.clf.eval()
+
+        # Get transformation
+        self.preprocess_transform = _get_preprocess_transform_cifar10()
+
+
+class Cifar100Classifier(ImageClassifier):
+    """CIFAR100 classifiers -- RepVGG (a2) -- based on PyTorch."""
+
+    def __init__(self, use_gpu=False):
+        """Initialises the image classifier."""
+        # Get class labels
+        self.class_idx = CIFAR100_LABEL_MAP
+
+        # Get the model
+        # https://github.com/huyvnphan/PyTorch_CIFAR10
+        clf = torch.hub.load(
+            'chenyaofo/pytorch-cifar-models',
+            'cifar100_repvgg_a2',
+            pretrained=True)
+
+        if use_gpu:
+            if CUDA_AVAILABLE:
+                clf = clf.to(DEVICE)
+                # clf.cuda()
+                predict_proba = self._predict_proba_gpu
+            else:
+                logger.warning('GPU was requested but it is not available. '
+                               'Using CPU instead.')
+                predict_proba = self._predict_proba_cpu
+        else:
+            predict_proba = self._predict_proba_cpu
+        self.predict_proba = predict_proba
+
+        self.clf = clf
+        self.clf.eval()
+
+        # Get transformation
+        self.preprocess_transform = _get_preprocess_transform_cifar100()