自己整理的一些tensorflow下ķeras实现的模型,可在Tensorflow2.X下运行
- AlexNet
- Darknet53
- DenseNet
- Dual_path_network
- GoogleNet
- MNasNet
- Resnet34
- Resnet50
- SEResNeXt
- VGG16
- Squeeze_Excite-Network
- MobileNetV3
- Efficientnet
- SE_HRNet
- ResNest
- FCN8S
- ICNet
- MiniNetv2
- PSPNet-ResNet50
- RAUNet-3D
- Refinenet
- Segnet
- Unet
- Unet_Xception_Resnetblock
- ResNextFPN
- Deeplabv2
- Deeplabv3+
- FastFCN
- HRNet
- ResUNet-a
- RCNN-UNet
- Attention Unet
- RCNN-Attention Unet
- UNet ++
1、UNet -U-Net:用于生物医学图像分割的卷积网络 https://arxiv.org/abs/1505.04597
2、RCNN-UNet-基于U-Net的递归残积卷积神经网络(R2U-Net)用于医学图像分割 https://arxiv.org/abs/1802.06955
3、Attention Unet -Attention U-Net:学习在哪里寻找胰腺 https://arxiv.org/abs/1804.03999
4、RCNN-Attention Unet -Attention R2U-Net:只需将两个最新的高级作品集成在一起(R2U-Net + Attention U-Net)
5、嵌套的UNet -UNet ++:用于医学图像分割的嵌套U-Net体系结构 https://arxiv.org/abs/1807.10165
Unet-Segmentation-Pytorch-Nest-of-Unets
不同点:我的实现初始滤波数32,原始为64。
- Focal_Tversky_loss
- C_Focal_loss
- B_Focal_loss
- LovaszSoftmax
- WeightedCCE
- jaccard_loss
- bce_jaccard_loss
- cce_jaccard_loss
- dice_loss
- bce_dice_loss
- cce_dice_loss
- iou_score
- jaccard_score
- f1_score
- f2_score
- dice_score
- gelu
- swish
- mish
- SE
- Res2Net
- Deformable_Conv
- FRN
- attention(PAM空间注意力和CAM通道注意力)
- BiFPN
- TCN(时间卷积网络——解决LSTM的并发问题)