|
- from __future__ import absolute_import
- from __future__ import print_function
- from __future__ import division
- import os
- import numpy as np
- from PIL import Image
- import torchvision.datasets as dst
-
- '''
- Modified from https://github.com/HobbitLong/RepDistiller/blob/master/dataset/cifar100.py
- '''
-
- class CIFAR10IdxSample(dst.CIFAR10):
- def __init__(self, root, train=True,
- transform=None, target_transform=None,
- download=False, n=4096, mode='exact', percent=1.0):
- super().__init__(root=root, train=train, download=download,
- transform=transform, target_transform=target_transform)
- self.n = n
- self.mode = mode
-
- num_classes = 10
- num_samples = len(self.data)
- labels = self.targets
-
- self.cls_positive = [[] for _ in range(num_classes)]
- for i in range(num_samples):
- self.cls_positive[labels[i]].append(i)
-
- self.cls_negative = [[] for _ in range(num_classes)]
- for i in range(num_classes):
- for j in range(num_classes):
- if j == i:
- continue
- self.cls_negative[i].extend(self.cls_positive[j])
-
- self.cls_positive = [np.asarray(self.cls_positive[i]) for i in range(num_classes)]
- self.cls_negative = [np.asarray(self.cls_negative[i]) for i in range(num_classes)]
-
- if 0 < percent < 1:
- num = int(len(self.cls_negative[0]) * percent)
- self.cls_negative = [np.random.permutation(self.cls_negative[i])[0:num]
- for i in range(num_classes)]
-
- self.cls_positive = np.asarray(self.cls_positive)
- self.cls_negative = np.asarray(self.cls_negative)
-
- def __getitem__(self, index):
- img, target = self.data[index], self.targets[index]
-
- img = Image.fromarray(img)
- if self.transform is not None:
- img = self.transform(img)
-
- if self.target_transform is not None:
- target = self.target_transform(target)
-
- if self.mode == 'exact':
- pos_idx = index
- elif self.mode == 'relax':
- pos_idx = np.random.choice(self.cls_positive[target], 1)[0]
- else:
- raise NotImplementedError(self.mode)
- replace = True if self.n > len(self.cls_negative[target]) else False
- neg_idx = np.random.choice(self.cls_negative[target], self.n, replace=replace)
- sample_idx = np.hstack((np.asarray([pos_idx]), neg_idx))
-
- return img, target, index, sample_idx
-
-
- class CIFAR100IdxSample(dst.CIFAR100):
- def __init__(self, root, train=True,
- transform=None, target_transform=None,
- download=False, n=4096, mode='exact', percent=1.0):
- super().__init__(root=root, train=train, download=download,
- transform=transform, target_transform=target_transform)
- self.n = n
- self.mode = mode
-
- num_classes = 100
- num_samples = len(self.data)
- labels = self.targets
-
- self.cls_positive = [[] for _ in range(num_classes)]
- for i in range(num_samples):
- self.cls_positive[labels[i]].append(i)
-
- self.cls_negative = [[] for _ in range(num_classes)]
- for i in range(num_classes):
- for j in range(num_classes):
- if j == i:
- continue
- self.cls_negative[i].extend(self.cls_positive[j])
-
- self.cls_positive = [np.asarray(self.cls_positive[i]) for i in range(num_classes)]
- self.cls_negative = [np.asarray(self.cls_negative[i]) for i in range(num_classes)]
-
- if 0 < percent < 1:
- num = int(len(self.cls_negative[0]) * percent)
- self.cls_negative = [np.random.permutation(self.cls_negative[i])[0:num]
- for i in range(num_classes)]
-
- self.cls_positive = np.asarray(self.cls_positive)
- self.cls_negative = np.asarray(self.cls_negative)
-
- def __getitem__(self, index):
- img, target = self.data[index], self.targets[index]
-
- img = Image.fromarray(img)
- if self.transform is not None:
- img = self.transform(img)
-
- if self.target_transform is not None:
- target = self.target_transform(target)
-
- if self.mode == 'exact':
- pos_idx = index
- elif self.mode == 'relax':
- pos_idx = np.random.choice(self.cls_positive[target], 1)[0]
- else:
- raise NotImplementedError(self.mode)
- replace = True if self.n > len(self.cls_negative[target]) else False
- neg_idx = np.random.choice(self.cls_negative[target], self.n, replace=replace)
- sample_idx = np.hstack((np.asarray([pos_idx]), neg_idx))
-
- return img, target, index, sample_idx
|