Backbone发展与语义分割网络发展

整理如下（按照arxiv上面时间线的预印版本来整理）:

Backbone（基础网络,也可以理解为分类网络）：

Backbone可以塞入UNET作为使用。

 

年代	网络名称与代码	论文名称
1989	LeNet	Backpropagation Applied to Handwritten Zip Code Recognition
1995	LeNet4, Boosted LeNet4, LeNet5	Comparison of learning algorithms for handwritten digit recognition
1998	LeNet4, Boosted LeNet4, LeNet5	GradientBased Learning Applied to Document Recognition
2012	AlexNet	ImageNet Classification with Deep Convolutional Neural Networks
2013-12-16	NiN	Network In Network
2014-9-4	VGG16-VGG19	VERY DEEP CONVOLUTIONAL NETWORKS FOR LARGE-SCALE IMAGE RECOGNITION
2014-9-17	Inception v1 (又称为GoogLeNet)	Going Deeper with Convolutions
2015-2-6	MSRANet	Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification
2015-12-2	Inception v2 (又称为GoogLeNet)	Rethinking the Inception Architecture for Computer Vision
2015-12-10	ResNet	Deep Residual Learning for Image Recognition
2016-2-23	Inception v4 (又称为GoogLeNet)	Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning
2016-2-24	SqueezeNet	SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and <0.5MB model size
2016-3-16	ResNet v2	Identity Mappings in Deep Residual Networks
2016-5-23	Wide ResNet	Wide Residual Networks
2016-8-25	DenseNet	Densely Connected Convolutional Networks
2016-10-7	Inception v3 (又称为GoogLeNet)	Xception: Deep Learning with Depthwise Separable Convolutions
2016-11-16	ResNext	Aggregated Residual Transformations for Deep Neural Networks
2017-4-17	MobileNet	MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications
2017-7-4	ShuffleNet	ShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile Devices
2017-7-6	DPNet	Dual Path Networks
2017-9-5	SENet: SE-ResNet，SE-ResNext	Squeeze-and-Excitation Networks
2017-10-26	Capsules	Dynamic Routing Between Capsules
2018-1-13	MobileNet v2	MobileNetV2: Inverted Residuals and Linear Bottlenecks
2018-5-23	SqueezeNext	SqueezeNext: Hardware-Aware Neural Network Design
2018-7-30	ShuffleNet V2	ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design
2018-7-31	NasNet	MnasNet: Platform-Aware Neural Architecture Search for Mobile
2019-1-24	AutoShuffleNet	AutoShuffleNet: Learning Permutation Matrices via an Exact Lipschitz Continuous Penalty in Deep Convolutional Neural Networks
2019-5-6	MobileNet v3	Searching for MobileNetV3

 

 

 

 

 

语义分割网络：

年代	网络名称与代码	论文名称
2013-11-11	RCNN	Rich feature hierarchies for accurate object detection and semantic segmentation
2014-6-18	SPP（目标检测）	Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition
2014-11-14	FCN	Fully Convolutional Networks for Semantic Segmentation
2014-12-22	DeepLab v1	SEMANTIC IMAGE SEGMENTATION WITH DEEP CONVOLUTIONAL NETS AND FULLY CONNECTED CRFS
2015-5-18	UNET	U-Net: Convolutional Networks for Biomedical Image Segmentation
2015-6-4	Faster R-CNN（里面提出了RPN）	Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
2015-6-8	YOLO v1	You Only Look Once: Unified, Real-Time Object Detection
2015-11-9	SegNet	Bayesian SegNet: Model Uncertainty in Deep Convolutional Encoder-Decoder Architectures for Scene Understanding
2015-12月	Fast R-CNN	Fast R-CNN
2015-12-8	SSD	SSD: Single Shot MultiBox Detector
2016-6-2	DeepLab v2	DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs
2016-11-20	RefineNet	RefineNet: Multi-Path Refinement Networks for High-Resolution Semantic Segmentation
2016-12-4	PSPNet	Pyramid Scene Parsing Network
2016-12-9	FPN	Feature Pyramid Networks for Object Detection
2016-12-25	YOLO v2	YOLO9000: Better, Faster, Stronger
2017-3-20	Mask-RCNN	Mask R-CNN
2017-6-13	DeepLab v3	Rethinking Atrous Convolution for Semantic Image Segmentation
2018-4-8	YOLO v3	YOLOv3: An Incremental Improvement

 

 

各种网络的实现框架可以参考[1],

各种网络的综述可以参考[2]

想看个大概的可以翻阅[3]

[4]的内容很有意思，可以看下：

 

[1]https://blog.csdn.net/helloworld_Fly/article/details/80306117

[2]https://arxiv.org/pdf/1704.06857.pdf

[3]https://blog.csdn.net/qq_20084101/article/details/80432960

[4]https://www.jiqizhixin.com/articles/092301