braincog.datasets package
Subpackages
- braincog.datasets.ESimagenet package
- braincog.datasets.NOmniglot package
Submodules
braincog.datasets.CUB2002011 module
- class braincog.datasets.CUB2002011.CUB2002011(root, train=True, transform=None, target_transform=None, download=False)
基类:
VisionDataset
CUB-200-2011 Dataset. :param root: Root directory of the dataset. :type root: string :param train: If True, creates dataset from training set, otherwise
creates from test set.
- 参数
transform (callable, optional) – A function/transform that takes in an PIL image and returns a transformed version. E.g,
transforms.RandomCrop
target_transform (callable, optional) – A function/transform that takes in the target and transforms it.
download (bool, optional) – If true, downloads the dataset from the internet and puts it in root directory. If dataset is already downloaded, it is not downloaded again.
- base_folder = 'CUB_200_2011/images'
- file_id = '1hbzc_P1FuxMkcabkgn9ZKinBwW683j45'
- filename = 'CUB_200_2011.tgz'
- tgz_md5 = '97eceeb196236b17998738112f37df78'
braincog.datasets.StanfordDogs module
- class braincog.datasets.StanfordDogs.StanfordDogs(root, train=True, transform=None, target_transform=None, download=False)
基类:
VisionDataset
Stanford Dogs Dataset. :param root: Root directory of the dataset. :type root: string :param train: If True, creates dataset from training set, otherwise
creates from test set.
- 参数
transform (callable, optional) – A function/transform that takes in an PIL image and returns a transformed version. E.g,
transforms.RandomCrop
target_transform (callable, optional) – A function/transform that takes in the target and transforms it.
download (bool, optional) – If true, downloads the dataset from the internet and puts it in root directory. If dataset is already downloaded, it is not downloaded again.
- download()
- download_url_prefix = 'http://vision.stanford.edu/aditya86/ImageNetDogs'
- load_split()
- stats()
braincog.datasets.TinyImageNet module
- class braincog.datasets.TinyImageNet.TinyImageNet(root, split='train', transform=None, target_transform=None, download=False)
基类:
VisionDataset
tiny-imageNet Dataset. :param root: Root directory of the dataset. :type root: string :param split: The dataset split, supports
train
, orval
. :type split: string, optional :param transform: A function/transform that takes in an PIL imageand returns a transformed version. E.g,
transforms.RandomCrop
- 参数
target_transform (callable, optional) – A function/transform that takes in the target and transforms it.
download (bool, optional) – If true, downloads the dataset from the internet and puts it in root directory. If dataset is already downloaded, it is not downloaded again.
- base_folder = 'tiny-imagenet-200/'
- filename = 'tiny-imagenet-200.zip'
- md5 = '90528d7ca1a48142e341f4ef8d21d0de'
- url = 'http://cs231n.stanford.edu/tiny-imagenet-200.zip'
- braincog.datasets.TinyImageNet.find_classes(class_file)
- braincog.datasets.TinyImageNet.make_dataset(root, base_folder, dirname, class_to_idx)
braincog.datasets.cut_mix module
- class braincog.datasets.cut_mix.CutMix(dataset, num_class, num_mix=1, beta=1.0, prob=1.0, indices=None, noise=0.0, vis=False, **kwargs)
基类:
Dataset
- class braincog.datasets.cut_mix.EventMix(dataset, num_class, num_mix=1, beta=1.0, prob=1.0, indices=None, noise=0.1, vis=False, gaussian_n=None, **kwargs)
基类:
Dataset
- braincog.datasets.cut_mix.GMM_mask(size, rat, n=None)
- braincog.datasets.cut_mix.GMM_mask_clip(size, rat)
- class braincog.datasets.cut_mix.MixUp(dataset, num_class, num_mix=1, beta=1.0, prob=1.0, indices=None, noise=0.0, vis=False, **kwargs)
基类:
Dataset
- braincog.datasets.cut_mix.calc_lam(x1, x2, bbt1, bbt2, bbx1, bbx2, bby1, bby2)
- braincog.datasets.cut_mix.calc_masked_lam(x1, x2, mask)
- braincog.datasets.cut_mix.calc_masked_lam_with_difference(x1, x2, mix, kernel_size=3)
- braincog.datasets.cut_mix.event_difference(x1, x2, kernel_size=3)
- braincog.datasets.cut_mix.onehot(size, target)
- braincog.datasets.cut_mix.rand_bbox(size, rat)
- braincog.datasets.cut_mix.rand_bbox_st(size, rat)
- braincog.datasets.cut_mix.rand_bbox_time(size, rat)
- braincog.datasets.cut_mix.spatio_mask(size, rat)
- braincog.datasets.cut_mix.st_mask(size, rat)
- braincog.datasets.cut_mix.temporal_mask(size, rat)
braincog.datasets.datasets module
- class braincog.datasets.datasets.MNISTData(data_path: str, batch_size: int, train_trans: Optional[Sequence[Module]] = None, test_trans: Optional[Sequence[Module]] = None, pin_memory: bool = True, drop_last: bool = True, shuffle: bool = True)
基类:
object
Load MNIST datesets.
- get_data_loaders()
- get_standard_data()
- braincog.datasets.datasets.build_dataset(is_train, img_size, dataset, path, same_da=False)
构建带有增强策略的数据集 :param is_train: 是否训练集 :param img_size: 输出图像尺寸 :param dataset: 数据集名称 :param path: 数据集路径 :param same_da: 为训练集使用测试集的增广方法 :return: 增强后的数据集
- braincog.datasets.datasets.build_transform(is_train, img_size)
构建数据增强, 适用于static data :param is_train: 是否训练集 :param img_size: 输出的图像尺寸 :return: 数据增强策略
- braincog.datasets.datasets.get_CUB2002011_data(batch_size, num_workers=8, same_da=False, *args, **kwargs)
- braincog.datasets.datasets.get_FGVCAircraft_data(batch_size, num_workers=8, same_da=False, *args, **kwargs)
- braincog.datasets.datasets.get_Flowers102_data(batch_size, num_workers=8, same_da=False, *args, **kwargs)
- braincog.datasets.datasets.get_NCALTECH101_data(batch_size, step, **kwargs)
获取NCaltech101数据 http://journal.frontiersin.org/Article/10.3389/fnins.2015.00437/abstract :param batch_size: batch size :param step: 仿真步长 :param kwargs: :return: (train loader, test loader, mixup_active, mixup_fn)
- braincog.datasets.datasets.get_NCARS_data(batch_size, step, **kwargs)
获取N-Cars数据 https://ieeexplore.ieee.org/document/8578284/ :param batch_size: batch size :param step: 仿真步长 :param kwargs: :return: (train loader, test loader, mixup_active, mixup_fn)
- braincog.datasets.datasets.get_StanfordCars_data(batch_size, num_workers=8, same_da=False, *args, **kwargs)
- braincog.datasets.datasets.get_StanfordDogs_data(batch_size, num_workers=8, same_da=False, *args, **kwargs)
- braincog.datasets.datasets.get_TinyImageNet_data(batch_size, num_workers=8, same_da=False, *args, **kwargs)
- braincog.datasets.datasets.get_cifar100_data(batch_size, num_workers=8, same_data=False, *args, **kwargs)
获取CIFAR100数据 https://www.cs.toronto.edu/~kriz/cifar.html :param batch_size: batch size :param kwargs: :return: (train loader, test loader, mixup_active, mixup_fn)
- braincog.datasets.datasets.get_cifar10_data(batch_size, num_workers=8, same_da=False, **kwargs)
- 获取CIFAR10数据
- 参数
batch_size – batch size
kwargs –
- 返回
(train loader, test loader, mixup_active, mixup_fn)
- braincog.datasets.datasets.get_dvsc10_data(batch_size, step, **kwargs)
获取DVS CIFAR10数据 http://journal.frontiersin.org/article/10.3389/fnins.2017.00309/full :param batch_size: batch size :param step: 仿真步长 :param kwargs: :return: (train loader, test loader, mixup_active, mixup_fn)
- braincog.datasets.datasets.get_dvsg_data(batch_size, step, **kwargs)
获取DVS Gesture数据 DOI: 10.1109/CVPR.2017.781 :param batch_size: batch size :param step: 仿真步长 :param kwargs: :return: (train loader, test loader, mixup_active, mixup_fn)
- braincog.datasets.datasets.get_esimnet_data(batch_size, step, **kwargs)
获取ES imagenet数据 DOI: 10.3389/fnins.2021.726582 :param batch_size: batch size :param step: 仿真步长,固定为8 :param reconstruct: 重构则时间步为1, 否则为8 :param kwargs: :return: (train loader, test loader, mixup_active, mixup_fn) :note: 没有自动下载, 下载及md5请参考spikingjelly, sampler默认为DistributedSampler
- braincog.datasets.datasets.get_fashion_data(batch_size, num_workers=8, same_da=False, **kwargs)
获取fashion MNIST数据 http://arxiv.org/abs/1708.07747 :param batch_size: batch size :param same_da: 为训练集使用测试集的增广方法 :param kwargs: :return: (train loader, test loader, mixup_active, mixup_fn)
- braincog.datasets.datasets.get_imnet_data(args, _logger, data_config, num_aug_splits, **kwargs)
获取ImageNet数据集 http://arxiv.org/abs/1409.0575 :param args: 其他的参数 :param _logger: 日志路径 :param data_config: 增强策略 :param num_aug_splits: 不同增强策略的数量 :param kwargs: :return: (train loader, test loader, mixup_active, mixup_fn)
- braincog.datasets.datasets.get_mnist_data(batch_size, num_workers=8, same_da=False, **kwargs)
获取MNIST数据 http://data.pymvpa.org/datasets/mnist/ :param batch_size: batch size :param same_da: 为训练集使用测试集的增广方法 :param kwargs: :return: (train loader, test loader, mixup_active, mixup_fn)
- braincog.datasets.datasets.get_nomni_data(batch_size, train_portion=1.0, **kwargs)
获取N-Omniglot数据 :param batch_size:batch的大小 :param data_mode:一共full nkks pair三种模式 :param frames_num:一个样本帧的个数 :param data_type:event frequency两种模式
- braincog.datasets.datasets.unpack_mix_param(args)
braincog.datasets.gen_input_signal module
braincog.datasets.rand_aug module
- braincog.datasets.rand_aug.CutoutAbs(x, v)
- braincog.datasets.rand_aug.CutoutTemporal(x, v)
- braincog.datasets.rand_aug.GaussianBlur(x, v)
- braincog.datasets.rand_aug.Identity(x, v)
- class braincog.datasets.rand_aug.RandAugment(n, m)
基类:
object
- braincog.datasets.rand_aug.Rotate(x, v)
- braincog.datasets.rand_aug.SaltAndPepperNoise(x, v)
- braincog.datasets.rand_aug.ShearX(x, v)
- braincog.datasets.rand_aug.ShearY(x, v)
- braincog.datasets.rand_aug.SpatioShift(x, v)
- braincog.datasets.rand_aug.TemporalShift(x, v)
- braincog.datasets.rand_aug.TranslateX(x, v)
- braincog.datasets.rand_aug.TranslateY(x, v)
- braincog.datasets.rand_aug.drop(x, v)
braincog.datasets.utils module
- braincog.datasets.utils.dvs_channel_check_expend(x)
检查是否存在DVS数据缺失, N-Car中有的数据会缺少一个通道 :param x: 输入的tensor :return: 补全之后的数据
- braincog.datasets.utils.rescale(x, factor=None)
数据放缩函数 :param x: 输入的tensor :param factor: 缩放因子 :return: 缩放后的数据
Module contents
- braincog.datasets.build_dataset(is_train, img_size, dataset, path, same_da=False)
构建带有增强策略的数据集 :param is_train: 是否训练集 :param img_size: 输出图像尺寸 :param dataset: 数据集名称 :param path: 数据集路径 :param same_da: 为训练集使用测试集的增广方法 :return: 增强后的数据集
- braincog.datasets.build_transform(is_train, img_size)
构建数据增强, 适用于static data :param is_train: 是否训练集 :param img_size: 输出的图像尺寸 :return: 数据增强策略
- braincog.datasets.dvs_channel_check_expend(x)
检查是否存在DVS数据缺失, N-Car中有的数据会缺少一个通道 :param x: 输入的tensor :return: 补全之后的数据
- braincog.datasets.get_NCALTECH101_data(batch_size, step, **kwargs)
获取NCaltech101数据 http://journal.frontiersin.org/Article/10.3389/fnins.2015.00437/abstract :param batch_size: batch size :param step: 仿真步长 :param kwargs: :return: (train loader, test loader, mixup_active, mixup_fn)
- braincog.datasets.get_NCARS_data(batch_size, step, **kwargs)
获取N-Cars数据 https://ieeexplore.ieee.org/document/8578284/ :param batch_size: batch size :param step: 仿真步长 :param kwargs: :return: (train loader, test loader, mixup_active, mixup_fn)
- braincog.datasets.get_cifar100_data(batch_size, num_workers=8, same_data=False, *args, **kwargs)
获取CIFAR100数据 https://www.cs.toronto.edu/~kriz/cifar.html :param batch_size: batch size :param kwargs: :return: (train loader, test loader, mixup_active, mixup_fn)
- braincog.datasets.get_cifar10_data(batch_size, num_workers=8, same_da=False, **kwargs)
- 获取CIFAR10数据
- 参数
batch_size – batch size
kwargs –
- 返回
(train loader, test loader, mixup_active, mixup_fn)
- braincog.datasets.get_dvsc10_data(batch_size, step, **kwargs)
获取DVS CIFAR10数据 http://journal.frontiersin.org/article/10.3389/fnins.2017.00309/full :param batch_size: batch size :param step: 仿真步长 :param kwargs: :return: (train loader, test loader, mixup_active, mixup_fn)
- braincog.datasets.get_dvsg_data(batch_size, step, **kwargs)
获取DVS Gesture数据 DOI: 10.1109/CVPR.2017.781 :param batch_size: batch size :param step: 仿真步长 :param kwargs: :return: (train loader, test loader, mixup_active, mixup_fn)
- braincog.datasets.get_fashion_data(batch_size, num_workers=8, same_da=False, **kwargs)
获取fashion MNIST数据 http://arxiv.org/abs/1708.07747 :param batch_size: batch size :param same_da: 为训练集使用测试集的增广方法 :param kwargs: :return: (train loader, test loader, mixup_active, mixup_fn)
- braincog.datasets.get_imnet_data(args, _logger, data_config, num_aug_splits, **kwargs)
获取ImageNet数据集 http://arxiv.org/abs/1409.0575 :param args: 其他的参数 :param _logger: 日志路径 :param data_config: 增强策略 :param num_aug_splits: 不同增强策略的数量 :param kwargs: :return: (train loader, test loader, mixup_active, mixup_fn)
- braincog.datasets.get_mnist_data(batch_size, num_workers=8, same_da=False, **kwargs)
获取MNIST数据 http://data.pymvpa.org/datasets/mnist/ :param batch_size: batch size :param same_da: 为训练集使用测试集的增广方法 :param kwargs: :return: (train loader, test loader, mixup_active, mixup_fn)
- braincog.datasets.get_nomni_data(batch_size, train_portion=1.0, **kwargs)
获取N-Omniglot数据 :param batch_size:batch的大小 :param data_mode:一共full nkks pair三种模式 :param frames_num:一个样本帧的个数 :param data_type:event frequency两种模式
- braincog.datasets.rescale(x, factor=None)
数据放缩函数 :param x: 输入的tensor :param factor: 缩放因子 :return: 缩放后的数据