魅力-语义分割模型

本库包含有常见的语义分割网络模型，例如FCN，U-Net，DeepLab系列，ResNet系列，ENet，MobileNet系列，PASNet，PSPNet，HRNet，ICNet，AttUNet，NesterUNet，SEUNet，VGG，R2UNet，SegNet等网络。包含有多种损失函数选择，多种学习率调整方式。

data

build:

------train:
------------images
------------labels
------val:
---------images
---------labels
------test:
----------images
----------labels
ccd:
------train:
------------images
------------labels
------val:
---------images
---------labels
------test:
----------images
----------labels
camvid:
------train:
------------images
------------labels
------val:
---------images
---------labels
------test:
----------images
----------labels

loss

包含有各种损失函数，例如

from losses.B_Focal_loss import focal_loss_binary
from losses.C_Focal_loss import focal_loss_multiclasses
from losses.Dice_loss import DiceLoss
from losses.BCE_Dice_loss import BCE_DiceLoss
from losses.CE_Dice_loss import CE_DiceLoss
from losses.Tversky_loss import TverskyLoss
from losses.Focal_Tversky_loss import FocalTverskyLoss
from losses.Weighted_Categorical_loss import Weighted_Categorical_CrossEntropy_Loss
from losses.Generalized_Dice_loss import GeneralizedDiceLoss
from losses.Jaccard_loss import JaccardLoss
from losses.BCE_Jaccard_Loss import BCE_JaccardLoss
from losses.CE_Jaccard_Loss import CE_JaccardLoss

metrics

'iou_score', 'jaccard_score', 'f1_score', 'f2_score',
'dice_score', 'get_f_score', 'get_iou_score', 'get_jaccard_score'

lr

CosineDecayScheduler
CosineDecayScheduler+
ReduceLROnPlateau
lr