Classification¶
Below you will find all the latest image classification models.
By convention, model names starting with lowercase are pretrained on imagenet while uppercase are not (vanilla). To load one of the pretrained
models with your own number of classes use the models.model_utils.get_model(...)
function and specify the name of the model
exactly like the pretrained model method name (e.g. if the method name reads pywick.models.classification.dpn.dualpath.dpn68
then use
dpn68 as the model name for models.model_utils.get_model(...)
.
Note: Since Pywick v0.6.5 we include 200+ models from rwightman’s repo which can be used by simply specifying the appropriate model name (all lowercase) in the yaml configuration file!
DualPathNet¶
PyTorch implementation of Dual Path Networks. Based on original MXNet implementation with many ideas from another PyTorch implementation.
This implementation is compatible with the pretrained weights from cypw’s MXNet implementation.
-
class
pywick.models.classification.dpn.dualpath.
DPN
(small=False, num_init_features=64, k_r=96, groups=32, b=False, k_sec=(3, 4, 20, 3), inc_sec=(16, 32, 24, 128), num_classes=1000, test_time_pool=False)[source]¶
-
pywick.models.classification.dpn.dualpath.
dpn68
(num_classes=1000, pretrained=False, test_time_pool=True)[source]¶ Pretrained DPN68 model
-
pywick.models.classification.dpn.dualpath.
dpn68b
(num_classes=1000, pretrained=False, test_time_pool=True)[source]¶ Pretrained DPN68b model
-
pywick.models.classification.dpn.dualpath.
dpn98
(num_classes=1000, pretrained=False, test_time_pool=True)[source]¶ Pretrained DPN98 model
FBResnet¶
Facebook implementation of ResNet
-
pywick.models.classification.fbresnet.
FBResNet18
(num_classes=1000)[source]¶ Constructs a ResNet-18 model.
- Args:
- num_classes
-
pywick.models.classification.fbresnet.
FBResNet34
(num_classes=1000)[source]¶ Constructs a ResNet-34 model.
- Args:
- num_classes
-
pywick.models.classification.fbresnet.
FBResNet50
(num_classes=1000)[source]¶ Constructs a ResNet-50 model.
- Args:
- num_classes
Inception_Resv2_wide¶
Inception Resnet V2 Wide implementation
InceptionResnetV2¶
InceptionResNetV2 model architecture from the “InceptionV4, Inception-ResNet…” paper.
InceptionV4¶
NASNet_mobile¶
NASNet Mobile following the paper: Learning Transferable Architectures for Scalable Image Recognition
PNASNnet¶
PNASNet-5 model architecture from the “Progressive Neural Architecture Search” paper.
Polynet¶
PolyNet architecture from the paper PolyNet: A Pursuit of Structural Diversity in Very Deep Networks.
Pyramid_Resnet¶
Implementation from paper: Deep Pyramidal Residual Networks. Not pretrained.
-
pywick.models.classification.pyramid_resnet.
PyResNet18
(pretrained=None, **kwargs)[source]¶ Not Pretrained
Resnet_preact¶
Preact_Resnet models. Not pretrained.
Resnet_swish¶
Resnet model combined with Swish activation function
-
class
pywick.models.classification.resnet_swish.
ResNet_swish
(block, layers, num_classes=1000)[source]¶
-
pywick.models.classification.resnet_swish.
ResNet18_swish
(pretrained=False, **kwargs)[source]¶ Constructs a ResNet-18 model. Not pretrained.
-
pywick.models.classification.resnet_swish.
ResNet34_swish
(pretrained=False, **kwargs)[source]¶ Constructs a ResNet-34 model. Not pretrained.
-
pywick.models.classification.resnet_swish.
ResNet50_swish
(pretrained=False, **kwargs)[source]¶ Constructs a ResNet-50 model. Not pretrained.
Resnext¶
Implementation of paper: Aggregated Residual Transformations for Deep Neural Networks.
-
pywick.models.classification.resnext.
resnext50_32x4d
(num_classes=1000, pretrained='imagenet')[source]¶ Pretrained Resnext50_32x4d model
SENet¶
SENet implementation as described in: Squeeze-and-Excitation Networks.
-
class
pywick.models.classification.senet.
SENet
(block, layers, groups, reduction, dropout_p=0.2, inplanes=128, input_3x3=True, downsample_kernel_size=3, downsample_padding=1, num_classes=1000)[source]¶
-
pywick.models.classification.senet.
senet154
(num_classes=1000, pretrained='imagenet')[source]¶ Pretrained SENet154 model
-
pywick.models.classification.senet.
se_resnet50
(num_classes=1000, pretrained='imagenet')[source]¶ Pretrained SEResNet50 model
-
pywick.models.classification.senet.
se_resnet101
(num_classes=1000, pretrained='imagenet')[source]¶ Pretrained SEResNet101 model
-
pywick.models.classification.senet.
se_resnet152
(num_classes=1000, pretrained='imagenet')[source]¶ Pretrained SEResNet152 model
WideResnet¶
Implementation of WideResNet as described in: Wide Residual Networks.
XCeption¶
Ported to pytorch thanks to [tstandley](https://github.com/tstandley/Xception-PyTorch)
@author: tstandley Adapted by cadene
Creates an Xception Model as defined in:
Francois Chollet Xception: Deep Learning with Depthwise Separable Convolutions.