Skip to content

Latest commit

 

History

History
182 lines (107 loc) · 8.12 KB

backbones.md

File metadata and controls

182 lines (107 loc) · 8.12 KB
Raw

paddleseg.models.backbone

The models subpackage contains backbones extracting features for sementic segmentation models.

ResNet_vd

ResNet_vd backbone from "Bag of Tricks for Image Classification with Convolutional Neural Networks"

CLASS paddleseg.models.backbones.Resnet_vd(layers=50, output_stride=None, multi_grid=(1, 1, 1), lr_mult_list=(0.1, 0.1, 0.2, 0.2), pretrained=None)

Args

  • layers (int, optional): The layers of ResNet_vd. The supported layers are [18, 34, 50, 101, 152, 200]. Default: 50.
  • output_stride (int, optional): The stride of output features compared to input images. It is 8 or 16. Default: 8.
  • multi_grid (tuple|list, optional): The grid of stage4. Defult: (1, 1, 1).
  • pretrained (str, optional): The path of pretrained model.

paddleseg.models.backbones.ResNet18_vd(**args)

Return a object of ResNet_vd class which layers is 18.

paddleseg.models.backbones.ResNet34_vd(**args)

Return a object of ResNet_vd class which layers is 34.

paddleseg.models.backbones.ResNet50_vd(**args)

Return a object of ResNet_vd class which layers is 50.

paddleseg.models.backbones.ResNet101_vd(**args)

Return a object of ResNet_vd class which layers is 101.

paddleseg.models.backbones.ResNet152_vd(**args)

Return a object of ResNet_vd class which layers is 152.

padddelseg.models.backbones.ResNet200_vd(**args)

Return a object of ResNet_vd class which layers is 200.

HRNet

HRNet backbone from "HRNet：Deep High-Resolution Representation Learning for Visual Recognition"

CLASS paddleseg.models.backbones.HRNet(pretrained=None, stage1_num_modules=1, stage1_num_blocks=(4,), stage1_num_channels=(64,), stage2_num_modules=1, stage2_num_blocks=(4, 4), stage2_num_channels=(18, 36), stage3_num_modules=4, stage3_num_blocks=(4, 4, 4), stage3_num_channels=(18, 36, 72), stage4_num_modules=3, stage4_num_blocks=(4, 4, 4, 4), stage4_num_channels=(18, 36, 72, 14), has_se=False, align_corners=False)

Args

  • pretrained (str, optional): The path of pretrained model.
  • stage1_num_modules (int, optional): Number of modules for stage1. Default 1.
  • stage1_num_blocks (list, optional): Number of blocks per module for stage1. Default (4,).
  • stage1_num_channels (list, optional): Number of channels per branch for stage1. Default (64,).
  • stage2_num_modules (int, optional): Number of modules for stage2. Default 1.
  • stage2_num_blocks (list, optional): Number of blocks per module for stage2. Default (4, 4).
  • stage2_num_channels (list, optional): Number of channels per branch for stage2. Default (18, 36).
  • stage3_num_modules (int, optional): Number of modules for stage3. Default 4.
  • stage3_num_blocks (list, optional): Number of blocks per module for stage3. Default (4, 4, 4).
  • stage3_num_channels (list, optional): Number of channels per branch for stage3. Default (18, 36, 72).
  • stage4_num_modules (int, optional): Number of modules for stage4. Default 3.
  • stage4_num_blocks (list, optional): Number of blocks per module for stage4. Default (4, 4, 4, 4).
  • stage4_num_channels (list, optional): Number of channels per branch for stage4. Default (18, 36, 72. 144).
  • has_se (bool, optional): Whether to use Squeeze-and-Excitation module. Default False.
  • align_corners (bool, optional): An argument of F.interpolate. It should be set to False when the feature size is even, e.g. 1024x512, otherwise it is True, e.g. 769x769. Default: False.

paddleseg.models.backbones.HRNet_W18_Small_V1(**kwargs)

Return a object of HRNet class which width is 18 and it is smaller than HRNet_W18_Small_V2.

paddleseg.models.backbones.HRNet_W18_Small_V2(**kwargs)

Return a object of HRNet class which width is 18 and it is smaller than HRNet_W18.

paddleseg.models.backbones.HRNet_W18(**kwargs)

Return a object of HRNet class which width is 18.

paddleseg.models.backbones.HRNet_W30(**kwargs)

Return a object of HRNet class which width is 30.

paddleseg.models.backbones.HRNet_W32(**kwargs)

Return a object of HRNet class which width is 32.

paddleseg.models.backbones.HRNet_W40(**kwargs)

Return a object of HRNet class which width is 40.

paddleseg.models.backbones.HRNet_W44(**kwargs)

Return a object of HRNet class which width is 44.

paddleseg.models.backbones.HRNet_W48(**kwargs)

Return a object of HRNet class which width is 48.

paddleseg.models.backbones.HRNet_W60(**kwargs)

Return a object of HRNet class which width is 60.

paddleseg.models.backbones.HRNet_W64(**kwargs)

Return a object of HRNet class which width is 64.

MobileNetV3

MobileNetV3 backbone from "Searching for MobileNetV3".

CLASS paddleseg.models.backbones.MobileNetV3(pretrained=None, scale=1.0, model_name="small", output_stride=None)

Args

  • pretrained (str, optional): The path of pretrained model.
  • scale (float, optional): The scale of channels . Default: 1.0.
  • model_name (str, optional): Model name. It determines the type of MobileNetV3. The value is 'small' or 'large'. Defualt: 'small'.
  • output_stride (int, optional): The stride of output features compared to input images. The value should be one of [2, 4, 8, 16, 32]. Default: None.

paddleseg.models.backbones.MobileNetV3_small_x0_35(**args)

Return a object of MobileNetV3 class which scale is 0.35 and model_name is small.

paddleseg.models.backbones.MobileNetV3_small_x0_5(**args)

Return a object of MobileNetV3 class which scale is 0.5 and model_name is small.

paddleseg.models.backbones.MobileNetV3_small_x0_75(**args)

Return a object of MobileNetV3 class which scale is 0.75 and model_name is small.

paddleseg.models.backbones.MobileNetV3_small_x1_0(**args)

Return a object of MobileNetV3 class which scale is 1.0 and model_name is small.

paddleseg.models.backbones.MobileNetV3_small_x1_25(**args)

Return a object of MobileNetV3 class which scale is 1.25 and model_name is small.

paddleseg.models.backbones.MobileNetV3_large_x0_35(**args)

Return a object of MobileNetV3 class which scale is 0.35 and model_name is large.

paddleseg.models.backbones.MobileNetV3_large_x0_5(**args)

Return a object of MobileNetV3 class which scale is 0.5 and model_name is large.

paddleseg.models.backbones.MobileNetV3_large_x0_75(**args)

Return a object of MobileNetV3 class which scale is 0.75 and model_name is large.

paddleseg.models.backbones.MobileNetV3_large_x1_0(**args)

Return a object of MobileNetV3 class which scale is 1.0 and model_name is large.

paddleseg.models.backbones.MobileNetV3_large_x1_25(**args)

Return a object of MobileNetV3 class which scale is 1.25 and model_name is large.

XceptionDeeplab

Xception backbone of DeepLabV3+ from "Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation"

CLASS paddleseg.models.backbones.XceptionDeeplab(backbone, pretrained=None, output_stride=16)

Args

  • backbone (str): Which type of Xception_DeepLab to select. It should be one of ('xception_41', 'xception_65', 'xception_71').
  • pretrained (str, optional): The path of pretrained model.
  • output_stride (int, optional): The stride of output features compared to input images. It is 8 or 16. Default: 16.

paddleseg.models.backbones.Xception41_deeplab(**args)

Return a object of XceptionDeeplab class which layers is 41.

paddleseg.models.backbones.Xception65_deeplab(**args)

Return a object of XceptionDeeplab class which layers is 65.

paddleseg.models.backbones.Xception71_deeplab(**args)

Return a object of XceptionDeeplab class which layers is 71.