夸领导关键点匹配
要看的⽂章：SpyNet [31], PWC-Net [38] and LiteFlowNet [14]，SelFlow(⾃监督)
能检测遮挡的光流：
UnFlow: Unsupervid learning of optical flow with a bidirectional census loss
Occlusion aware unsupervid learning of optical flow
Unsupervid learning of multi-frame optical flow with occlusions
Ddflow: Learning optical flow with unlabeled data distillation
Back to basics:Unsupervid learning of optical flow via brightness constancy and motion smoothness
Unsupervid deep learning for optical flow estimation
. Unsupervid monocular depth estimation with left-right consistency.
月偏旁的字有哪些Optical flow estimation with channel constancy. 2014ECCV
Image quality asssment:from error visibility to structural similarity. 2004
Learning den correspondence via 3d-guided cycle consistency 2016 CVPR
Convolutional neural network architecture for geometric matching CVPR 2017
SIFT+RANSAC的进化版：
更好的特征描述：Superpoint: Self-supervid interest point detection and description. 2018
Geometric image correspondence verification by den pixel matching 2019
Geodesc:Learning local descriptors by integrating geometry constraints. 2018
Working hard to know your neighbor’s margins: Local descriptor learning loss NIPS2017
R2d2: Reliable and repeatable detector and descripto NIPS 2019
L2-net: Deep learning of discriminative patch descriptor in euclidean space CVPR 2017
RANSAC:Learning two-view correspondences and geometry using order-aware network ICCV 2019
Deep fundamental matrix estimation ECCV 2018
Pointnet: Deep learning on point ts for 3d classification and gmentation CVPR 2017
Neural nearest neighbors networks NIPS 2018
光流：
Flow Fields: Den Correspondence Fields for Highly Accurate Large Displacement Optical Flow Estimation ⼤范围光流场估计2015ICCV
使⽤局部特征进⾏光流匹配：Large displacement optical flow CVPR 2009
: Epicflow: Edge-prerving interpolation of correspondences for optical flow
Flow fields: Den correspondence fields for highly accurate large displacement optical flow estimation
: Efficient coar-to-fine patchmatch for large displacement optical flow
超越视觉相似性的光流：SIFTFlow,FlowWeb: Joint Image Set Alignment by
Weaving Consistent, Pixel-wi Correspondences
Do Convnets Learn Correspondence?
Universal correspondence network婚礼开场白主持词
Neighbourhood connsus networks
深度光流：Geonet: Unsupervid learning of den depth, optical flow and camera po
Learning correspondence from the cycle-consistency of time CVPR 2019
: Pwc-net: Cnns for optical flow using pyramid, warping, and cost volume CVPR 2018
Dgc-net: Den geometric correspondence network
Geometric image correspondence verification by den pixel matching
making convolutional networks shift-invariant again
1、⽴体匹配算法主要可分为两⼤类：基于局部约束和基于全局约束的⽴体匹配算法．
（⼀）基于全局约束的⽴体匹配算法：在本质上属于优化算法，它是将⽴体匹配问题转化为寻找全局能量函数的最优化问题，其代表算法主要有图割算法、置信度传播算法和协同优化算法等．全局算法能够获得较低的总误匹配率，但算法复杂度较⾼，很难满⾜实时的需求，不利于在实际⼯程中使⽤．葡萄的作文
（⼆）基于局部约束的⽴体匹配算法：主要是利⽤匹配点周围的局部信息进⾏计算，由于其涉及到的信息量较少，匹配时间较短，因此受到了⼴泛关注，其代表算法主要有 SAD、SSD、NCC等
1.SuperGlue: Learning Feature Matching with Graph Neural Networks（2020）
学困生辅导记录表
⽹络结构：
代码：
（1）keypoint encoder
对应于⽂中
lf.kenc = fig['descriptor_dim'], lf.config['keypoint_encoder'])
def__init__(lf, feature_dim, layers):
其中，MLP是多个1维的卷积，类似于FC，从通道3（x,y,c）依次变为通道[32,64,128,256]，再变为256 descriptor_dim。
layers.append( nn.Conv1d(channels[i -1], channels[i], kernel_size=1, bias=True))
desc0 = desc0 + lf.kenc(kpts0, data['scores0'])
def forward(lf, kpts, scores):
inputs =[anspo(1,2), scores.unsqueeze(1)]
der(torch.cat(inputs, dim=1))
2. Graph Neural Network with multiple lf and cross-attention layers
默认值：lf.config['descriptor_dim']=256，lf.config['GNN_layers']=['lf','cross']*9,
< = AttentionalGNN( lf.config['descriptor_dim'], lf.config['GNN_layers'])
lf.layers = nn.ModuleList([ AttentionalPropagation(feature_dim,4)for _ in range(len(layer_names))])#len(layer_names)=18，
AttentionalPropagation(feature_dim,4)
lf.attn = MultiHeadedAttention(num_heads, feature_dim)#MultiHeadedAttention(4, 256)
lf.dim = d_model // num_heads #256//4=64
lf.num_heads = num_heads #4
< = nn.Conv1d(d_model, d_model, kernel_size=1)#nn.Conv1d(256, 256, kernel_size=1)
lf.proj = nn.ModuleList([)for _ in range(3)])
forward的时候：def forward(lf, query, key, value):
放射性污染query, key, value =[l(x).view(batch_dim, lf.dim, lf.num_heads,-1)for l, x in zip(lf.proj,(query, key, value))]#lf.merge(
lf.mlp = MLP([feature_dim*2, feature_dim*2, feature_dim])#MLP([256*2, 256*2, 256])
desc0, desc1 = lf.gnn(desc0, desc1)
2.Flow2Stereo: Effective Self-Supervid Learning of Optical Flow and Stereo Matching
我的阳光牧场
采⽤⾃监督的⽅式，能够使⽤两个时间(t,t+1)和两个视⾓的4张图像，估计出图像间的光流和⽴体视差。
（1）⾸先，计算两个时间(t,t+1)和两个视⾓的4张图像，他们之间所满⾜的四边形限制和三⾓形限制
（2）采⽤两阶段的策略，teacher model和student model，其中teacher model阶段的损失函数为Photometric loss(基于：对于没有遮挡的像素的亮度⼀致性假设)，三⾓形限制和四边形限制；student model阶段的损失函数为⾃监督损失（基于teacher model预测的光流和confidence map）
四边形限制和三⾓形限制的计算：
两阶段的⽹络：
3. Ddflow: Learning optical flow with unlabeled data distillation
(1)在⽆监督学习中，使⽤亮度⼀致性构建损失函数是合理的，但是这条法则并不适⽤于有遮挡的像素
（2）所以，⾸先训练teacher model。其中，在训练teacher model时可以根据前项光流和后项光流的不⼀致性识别被遮挡的像素（这个假设不是完全成⽴的）。训练teacher model时只使⽤了亮度⼀致性损失
（3）训练完teacher model后，训练student model。对于有遮挡的地⽅，采⽤teacher model的预测结果作为label，计算损失。对于没有遮挡的地⽅，使⽤亮度⼀致性损失。
5.UnFlow: Unsupervid Learning of Optical Flow with a Bidirectional Census Loss
（1）定义被遮挡的区域：因为遮挡区域不满⾜亮度⼀致性假设，所以单独处理被遮挡的区域。我们假设对于没有遮挡的区域，前向
flow(I1->I2)应该是后向flow(I2->I1)的inver,⽽对于遮挡的区域，即为此条假设不成⽴的地⽅：
（2）对于没有遮挡的区域，⾃监督的损失由3部分组成：亮度⼀致性损失ED，光流平滑性ES(smooth)，前后向光流的⼀致相反性EC.对于被遮挡的区域，只计算光流平滑性损失。
桑椹怎么洗
