qualia2.vision package

Submodules

qualia2.vision.alexnet module

class qualia2.vision.alexnet.AlexNet(pretrained=False)

Bases: qualia2.nn.modules.module.Module

Args:

pretrained (bool): if true, load a pretrained weights

forward(x)

qualia2.vision.densenet module

class qualia2.vision.densenet.DenseBlock(num_layers, num_input_features, bn_size, growth_rate, drop_rate)

Bases: qualia2.nn.modules.module.Module

forward(init_features)

class qualia2.vision.densenet.DenseLayer(num_input_features, growth_rate, bn_size, drop_rate)

Bases: qualia2.nn.modules.module.Module

forward(*prev_features)

class qualia2.vision.densenet.DenseNet(growth_rate=32, block_config=(6, 12, 24, 16), num_init_features=64, bn_size=4, drop_rate=0, num_classes=1000, pretrained=False)

Bases: qualia2.nn.modules.module.Module

Densely Connected Convolutional Networks Args:

growth_rate (int): how many filters to add each layer (k in paper) block_config (list of 4 ints): how many layers in each pooling block num_init_features (int): the number of filters to learn in the first convolution layer bn_size (int): multiplicative factor for number of bottle neck layers (i.e. bn_size * k features in the bottleneck layer) drop_rate (float): dropout rate after each dense layer num_classes (int): number of classification classes

classmethod densenet121(pretrained=False)

classmethod densenet161(pretrained=False)

classmethod densenet169(pretrained=False)

classmethod densenet201(pretrained=False)

forward(x)

qualia2.vision.densenet.DenseNet121(pretrained=False)

qualia2.vision.densenet.DenseNet161(pretrained=False)

qualia2.vision.densenet.DenseNet169(pretrained=False)

qualia2.vision.densenet.DenseNet201(pretrained=False)

qualia2.vision.densenet.transition(num_input_features, num_output_features)

qualia2.vision.imagenet_labels module

qualia2.vision.openpose module

class qualia2.vision.openpose.OpenPoseBody(pretrained=False)

Bases: qualia2.nn.modules.module.Module

static create_block(block)

forward(x)

class qualia2.vision.openpose.OpenPoseHand(pretrained=False)

Bases: qualia2.nn.modules.module.Module

static create_block(block)

forward(x)

qualia2.vision.resnet module

class qualia2.vision.resnet.Basic(inplanes, planes, stride=1, downsample=None, base_width=64, dilation=1, norm_layer=<class 'qualia2.nn.modules.normalize.BatchNorm2d'>)

Bases: qualia2.nn.modules.module.Module

expansion = 1

forward(x)

class qualia2.vision.resnet.Bottleneck(inplanes, planes, stride=1, downsample=None, base_width=64, dilation=1, norm_layer=<class 'qualia2.nn.modules.normalize.BatchNorm2d'>)

Bases: qualia2.nn.modules.module.Module

expansion = 4

forward(x)

class qualia2.vision.resnet.ResNet(block, layers, num_classes=1000, zero_init_residual=False, replace_stride_with_dilation=[False, False, False], norm_layer=<class 'qualia2.nn.modules.normalize.BatchNorm2d'>, pretrained=False)

Bases: qualia2.nn.modules.module.Module

Args: block (Module): Basic Block to create layers layers (list of int): config of layers num_classes (int): size of output classes zero_init_residual (bool): Zero-initialize the last BN in each residual branch, so that the residual branch starts with zeros, and each residual block behaves like an identity.

This improves the model by 0.2~0.3% according to https://arxiv.org/abs/1706.02677

replace_stride_with_dilation (list of bool): each element in the list indicates if we should replace the 2x2 stride with a dilated convolution

forward(x)

classmethod resnet101(pretrained=False)

classmethod resnet152(pretrained=False)

classmethod resnet18(pretrained=False)

classmethod resnet34(pretrained=False)

classmethod resnet50(pretrained=False)

qualia2.vision.resnet.ResNet101(pretrained=False)

qualia2.vision.resnet.ResNet152(pretrained=False)

qualia2.vision.resnet.ResNet18(pretrained=False)

qualia2.vision.resnet.ResNet34(pretrained=False)

qualia2.vision.resnet.ResNet50(pretrained=False)

qualia2.vision.resnet.conv1x1(in_planes, out_planes, stride=1)

1x1 convolution

qualia2.vision.resnet.conv3x3(in_planes, out_planes, stride=1, dilation=1)

3x3 convolution with padding

qualia2.vision.squeezenet module

class qualia2.vision.squeezenet.Fire(inplanes, squeeze_planes, expand1x1_planes, expand3x3_planes)

Bases: qualia2.nn.modules.module.Module

forward(x)

class qualia2.vision.squeezenet.SqueezeNet(pretrained=False)

Bases: qualia2.nn.modules.module.Module

forward(x)

qualia2.vision.transforms module

class qualia2.vision.transforms.CenterCrop(size)

Bases: object

Crops the given PIL Image at the center. Args:

size (int): Desired output size of the crop.

class qualia2.vision.transforms.Compose(transforms)

Bases: object

Composes several transforms together. Args:

transforms (list): list of transforms

class qualia2.vision.transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])

Bases: object

Normalize a tensor image with mean and standard deviation.

class qualia2.vision.transforms.Resize(size, interpolation=2)

Bases: object

Resize the input PIL Image to the given size. Args:

size (int): Desired output size. (size * height / width, size) interpolation (int): Desired interpolation. Default is PIL.Image.BILINEAR

class qualia2.vision.transforms.ToPIL

Bases: object

Convert Tensor to PIL Image.

class qualia2.vision.transforms.ToTensor

Bases: object

Convert a PIL Image to Tensor.

qualia2.vision.vgg module

class qualia2.vision.vgg.VGG(ver, pretrained=False, batch_norm=False)

Bases: qualia2.nn.modules.module.Module

Base class of VGG

Args:

features (Module): feanture Module cfg (int): model config pretrained (bool): if true, load a pretrained weights

static create_layers(ver, batch_norm=False)

forward(x)

classmethod vgg11(pretrained=False)

classmethod vgg11_bn(pretrained=False)

classmethod vgg13(pretrained=False)

classmethod vgg13_bn(pretrained=False)

classmethod vgg16(pretrained=False)

classmethod vgg16_bn(pretrained=False)

classmethod vgg19(pretrained=False)

classmethod vgg19_bn(pretrained=False)

qualia2.vision.vgg.VGG11(pretrained=False)

qualia2.vision.vgg.VGG11_bn(pretrained=False)

qualia2.vision.vgg.VGG13(pretrained=False)

qualia2.vision.vgg.VGG13_bn(pretrained=False)

qualia2.vision.vgg.VGG16(pretrained=False)

qualia2.vision.vgg.VGG16_bn(pretrained=False)

qualia2.vision.vgg.VGG19(pretrained=False)

qualia2.vision.vgg.VGG19_bn(pretrained=False)

Module contents