MCPcopy Create free account

hub / github.com/BBuf/Keras-Semantic-Segmentation / functions

Functions163 in github.com/BBuf/Keras-Semantic-Segmentation

⨍ Functions 163◇ Types & classes 4
↓ 17 callersFunctionconv
(x, outsize, kernel_size, strides_=1, padding_='same', activation=None)
Models/HRNet.py:8
↓ 15 callersFunctionconv_block_nested
(x, mid_ch, out_ch, kernel_size=3, padding='same')
Models/NestedUNet.py:7
↓ 13 callersFunctiondepthwise_conv_block
(inputs, pointwise_conv_filters, alpha, depth_multiplier=1, strides=(1, 1), block_id=1)
Models/MobileNetFCN8.py:18
↓ 12 callersFunctionresidual_block
(blockInput, num_filters=16)
Models/UNet_Xception_ResNetBlock.py:14
↓ 11 callersFunctionbottleneck
(inp, output, internal_scale=4, asymmetric=0, dilated=0, downsample=False, dropout_rate=0.1)
Models/ENet.py:19
↓ 11 callersFunctiondepthwise_conv_block
(inputs, pointwise_conv_filters, alpha, depth_multiplier=1, strides=(1, 1), block_id=1)
Models/MobileNetUnet.py:17
↓ 9 callersFunctionRRCNN_block
(input, channel, t=2)
Models/R2UNet.py:34
↓ 9 callersFunctionSEModule
(input, ratio, out_dim)
Models/SEUNet.py:6
↓ 9 callersFunctionconv_block
(input, filters)
Models/AttUNet.py:6
↓ 8 callersFunctionconv_block
(input, filters)
Models/Unet.py:7
↓ 7 callersFunctionResNet
构造resnet Args: x: block_type: 根据不同的类型，选择BasicBlock或BottlenectBlock layers_repeat: 每一层重复的次数 class_dim: 最后Dense
Models/ACNet.py:145
↓ 5 callersFunctionConvBNLayer
卷积+归一化层 Args: x: 输入张量 out_channels: 输出通道数 kernel_size: 卷积核大小 stride: 步长 dilation: 膨胀系数，默认为1 a
Models/ACNet.py:4
↓ 5 callersFunctionRRCNN_block
(input, channel, t=2)
Models/R2AttUNet.py:46
↓ 5 callersFunctionde_bottleneck
(encoder, output, upsample=False, reverse_module=False)
Models/ENet.py:92
↓ 4 callersFunctionAttention_block
(input1, input2, filters)
Models/AttUNet.py:22
↓ 4 callersFunctionAttention_block
(input1, input2, filters)
Models/R2AttUNet.py:23
↓ 4 callersFunctionBottleneck
(x, size, downsampe=False)
Models/HRNet.py:13
↓ 4 callersFunctionConvBnLayer
(x, oup, kernel_size, stride, padding='valid')
Models/GhostNet.py:24
↓ 4 callersFunction_make_divisible
This function is taken from the original tf repo. It ensures that all layers have a channel number that is divisible by 8 It can be seen
Models/GhostNet.py:8
↓ 4 callersMethodbuildActivation
(self, name, activation)
tools/keras2Msnh/MsnhBuilder.py:31
↓ 4 callersFunctionbuild_model
(name, num_classes, input_height, input_width)
Models/__init__.py:56
↓ 4 callersFunctionconv_block
(input, filters)
Models/R2AttUNet.py:7
↓ 4 callersFunctionmk
(dir)
utils/prepare_helen_small.py:20
↓ 4 callersFunctionset_padding
(config_keras, input_shape, config_caffe)
tools/keras2caffe/convert.py:8
↓ 4 callersFunctionup_conv
(input, filters)
Models/Unet.py:16
↓ 4 callersFunctionup_conv
(input, filters)
Models/R2UNet.py:16
↓ 4 callersFunctionup_conv
(input, filters)
Models/AttUNet.py:15
↓ 4 callersFunctionup_conv
(input, filters)
Models/R2AttUNet.py:16
↓ 3 callersFunctionACBlock
ACBlock搭建，这里没有采取源代码的Crop，而是直接通过same填充 Args: x: 输入张量 out_channels: 输出通道 kernel_size: 卷积核大小 stride: 步长
Models/ACNet.py:27
↓ 3 callersFunctionConvBNReLU
Conv + BN + RELU6 Args: x: 输入张量 out_planes: 输出通道数 kernel_size: 卷积核大小 stride: 步长 groups: 分组，当=1则为普通卷积，
Models/MobileNext.py:32
↓ 3 callersFunctionDepthWiseConv
(x, kernel_size=3, stride=1, depth_multiplier=1, padding='same', relu=False)
Models/GhostNet.py:40
↓ 3 callersMethodbuildSoftmax
(self, name)
tools/keras2Msnh/MsnhBuilder.py:49
↓ 3 callersFunctioncsse_block
Implementation of Concurrent Spatial and Channel ‘Squeeze & Excitation’ in Fully Convolutional Networks https://arxiv.org/abs/1803.02579
Models/scSEUnet.py:32
↓ 3 callersFunctionstage
(x, num_modules, channels, multi_scale_output=True)
Models/HRNet.py:136
↓ 3 callersFunctiontransition_layer
(x, in_channels, out_channels)
Models/HRNet.py:66
↓ 2 callersFunctionCutOutLayer
在通道维上截取到 oup 的张量 Args: input_tensor: 输入张量 oup: 输出通道数 Returns:
Models/MobileNext.py:58
↓ 2 callersFunctionGhostModule
(x, oup, kernel_size=1, ratio=2, dw_size=3, stride=1, relu=True)
Models/GhostNet.py:49
↓ 2 callersFunctionHighResolutionModule
(x, channels, multi_scale_output=True)
Models/HRNet.py:129
↓ 2 callersFunctionRecurrent_block
(input, channel, t=2)
Models/R2UNet.py:23
↓ 2 callersFunctionRecurrent_block
(input, channel, t=2)
Models/R2AttUNet.py:35
↓ 2 callersFunction_make_divisible
This function is taken from the original tf repo. It ensures that all layers have a channel number that is divisible by 8 It can be seen
Models/MobileNext.py:12
↓ 2 callersMethodbuildRoute
(self, name, layers, addModel)
tools/keras2Msnh/MsnhBuilder.py:98
↓ 2 callersFunctionconvolution_block
(x, filters, size, strides=(1,1), padding='same', activation=True)
Models/UNet_Xception_ResNetBlock.py:7
↓ 2 callersFunctioncv2_letterbox_image
(image, dst_size)
utils/utils.py:13
↓ 2 callersFunctiongetPadding
(config_keras, input_shape)
tools/keras2Msnh/Keras2Msnh.py:6
↓ 2 callersMethodget_config
(self)
Models/DeepLabV3+.py:34
↓ 2 callersFunctionget_f_score
参数: class_weights: 1. or list of class weights, len(weights) = C smooth: value to avoid division by zero beta: f-scor
metrics/metrics.py:110
↓ 2 callersFunctionget_flops
(model)
utils/utils.py:28
↓ 2 callersFunctionmk_if_not_exits
(dir)
utils/utils.py:8
↓ 2 callersFunctionxception_downsample_block
(x, channels, top_relu=False)
Models/DeepLabV3+.py:40
↓ 1 callersFunctionBCE_DiceLoss
()
losses/BCE_Dice_loss.py:11
↓ 1 callersFunctionBCE_JaccardLoss
()
losses/BCE_Jaccard_Loss.py:27
↓ 1 callersFunctionBasicBlock
(x, size, downsampe=False)
Models/HRNet.py:37
↓ 1 callersFunctionCE_DiceLoss
()
losses/CE_Dice_loss.py:24
↓ 1 callersFunctionCE_JaccardLoss
()
losses/CE_Jaccard_Loss.py:27
↓ 1 callersFunctionCutInLayer
在通道维上截取从 oup开始到后面的张量 Args: input_tensor: 输入张量 oup: 输出通道数 Returns:
Models/MobileNext.py:71
↓ 1 callersFunctionDiceLoss
()
losses/Dice_loss.py:15
↓ 1 callersFunctionDice_Loss
(y_true, y_pred)
losses/CE_Dice_loss.py:16
↓ 1 callersFunctionDice_coeff
(y_true, y_pred)
losses/CE_Dice_loss.py:7
↓ 1 callersFunctionFCN8_helper
(nClasses, input_height=224, input_width=224)
Models/FCN8.py:8
↓ 1 callersFunctionFocalTverskyLoss
()
losses/Focal_Tversky_loss.py:21
↓ 1 callersFunctionGeneralizedDiceLoss
()
losses/Generalized_Dice_loss.py:22
↓ 1 callersFunctionGhostBottleneck
(x, hidden_dim, oup, kernel_size, stride, use_se)
Models/GhostNet.py:69
↓ 1 callersFunctionJaccardLoss
()
losses/Jaccard_loss.py:58
↓ 1 callersFunctionSELayer
(x, reduction=4)
Models/GhostNet.py:30
↓ 1 callersFunctionSandGlass
SandGlass 模块 Args: x: 输入张量 oup: 输出通道数 stride: 步长 expand_ratio: 扩张系数 identity_tensor_multiplier: 区间在0-
Models/MobileNext.py:84
↓ 1 callersFunctionTversky
(y_true, y_pred, smooth=1e-6)
losses/Tversky_loss.py:7
↓ 1 callersFunctionTversky
(y_true, y_pred, smooth=1e-6)
losses/Focal_Tversky_loss.py:7
↓ 1 callersFunctionTverskyLoss
()
losses/Tversky_loss.py:19
↓ 1 callersFunctionWeighted_Categorical_CrossEntropy_Loss
(dataset)
losses/Weighted_Categorical_loss.py:13
↓ 1 callersFunction_dice_coef_loss
(y_true, y_pred)
losses/Dice_loss.py:12
↓ 1 callersFunctionaspp
(x, input_shape, out_stride)
Models/DeepLabV3+.py:104
↓ 1 callersFunctionbranches
(x, block_num, channels)
Models/HRNet.py:90
↓ 1 callersMethodbuildBatchNorm
(self, name)
tools/keras2Msnh/MsnhBuilder.py:56
↓ 1 callersMethodbuildConfig
(self,shape)
tools/keras2Msnh/MsnhBuilder.py:7
↓ 1 callersMethodbuildConnect
(self, name, output)
tools/keras2Msnh/MsnhBuilder.py:84
↓ 1 callersMethodbuildConv2d
(self, name, filters, kSizeX, kSizeY, paddingX, paddingY, strideX, strideY, dilationX, dilationY, groups, useB
tools/keras2Msnh/MsnhBuilder.py:14
↓ 1 callersMethodbuildGlobalAvgPooling
(self, name)
tools/keras2Msnh/MsnhBuilder.py:62
↓ 1 callersMethodbuildPadding
(self, name, top, down, left, right)
tools/keras2Msnh/MsnhBuilder.py:106
↓ 1 callersMethodbuildPooling
(self, name, type, kSizeX, kSizeY, strideX, strideY, paddingX, paddingY)
tools/keras2Msnh/MsnhBuilder.py:68
↓ 1 callersMethodbuildUpsample2D
(self, name, stride)
tools/keras2Msnh/MsnhBuilder.py:91
↓ 1 callersFunctionconv_block
(inputs, filters, alpha, kernel=(3, 3), strides=(1, 1))
Models/MobileNetFCN8.py:11
↓ 1 callersFunctionconv_block
(inputs, filters, alpha, kernel=(3, 3), strides=(1, 1))
Models/MobileNetUnet.py:10
↓ 1 callersFunctioncse_block
(prevlayer, prefix)
Models/scSEUnet.py:8
↓ 1 callersFunctiondata_num
(train_path, mask_path)
albumentations.py:24
↓ 1 callersFunctionde_build
(encoder, nc)
Models/ENet.py:123
↓ 1 callersFunctiondice_coef
Dice = (2*|X & Y|)/ (|X|+ |Y|) = 2*sum(|A*B|)/(sum(A^2)+sum(B^2)) ref: https://arxiv.org/pdf/1606.04797v1.pdf
losses/Dice_loss.py:3
↓ 1 callersFunctionen_build
(inp, dropout_rate=0.01)
Models/ENet.py:69
↓ 1 callersFunctionf_score
(gt, pr, class_weights=1, beta=1, smooth=SMOOTH, per_image=True, threshold=None)
metrics/metrics.py:62
↓ 1 callersFunctionfocal_loss_binary
(gamma=2., alpha=.25)
losses/B_Focal_loss.py:5
↓ 1 callersFunctionfocal_loss_multiclasses
(dataset, gamma=2., e=0.1)
losses/C_Focal_loss.py:10
↓ 1 callersFunctionfuse_layers
(x, channels, multi_scale_output=True)
Models/HRNet.py:100
↓ 1 callersFunctiongeneralized_dice_coeff
(y_true, y_pred)
losses/Generalized_Dice_loss.py:4
↓ 1 callersFunctiongetImage
(path, width, height, imgNorm="sub_mean", resize_op=1)
data.py:10
↓ 1 callersFunctiongetLable
(path, n_classes, width, height, resize_op)
data.py:31
↓ 1 callersFunctioninitial_block
(inp, nb_filter=13, nb_row=3, nb_col=3, strides=(2, 2))
Models/ENet.py:13
↓ 1 callersFunctioniou_score
参数： gt: ground truth 4D keras tensor (B, H, W, C) pr: prediction 4D keras tensor (B, H, W, C) class_weights: 1. or li
metrics/metrics.py:11
↓ 1 callersFunctionkeras2Msnh
(keras_model, msnhnet_file, msnhbin_file)
tools/keras2Msnh/Keras2Msnh.py:35
↓ 1 callersFunctionlayer1
(x)
Models/HRNet.py:57
↓ 1 callersFunctionmain
()
albumentations.py:38
next →1–100 of 163, ranked by callers