### Deep Learning of Graph Matching论文解读

SIGAI特约作者

$$J(x)=_x^{max}X^{T}Mx\\s.t.X1=1,x^{T}1\leq 1,x\in (0,1)^{m \times n}$$

$$M_{p}=U^{1}U^{2T}$$

$$X_{c}=[F_{i}^{1}|F_{j}^{1}],Y_{c}=[[F_{i}^{2}|F_{j}^{2}]$$

$$M_{e}=X\Lambda Y^{T}$$

$$v_{k+1}=\frac{Mv_{k}}{\left \| Mv_{k} \right \|}$$

$$S_{K+1}=S_{k}[1_{n}^{T}S_{k}]^{-1},S_{K+2}=[S_{k+1}1_{m}]^{-1}S_{k+1}$$

$$L(d)=\sum_{i}\varnothing (d_{i}-d_{i}^{gt}), \varnothing (x)=\sqrt{X^{T}X+\Theta }$$

Deep Learning of Graph Matching这篇工作首次将端到端的深度学习与图匹配问题结合，在学术界已经引起了不小的关注。结合机器学习，尤其是深度学习，提升传统计算机视觉算法的精度，是学术界发展的趋势之一。基于这篇工作，后续可能可以从主干网络、匹配函数、图结构学习、图匹配求解入手，进一步提高深度学习图匹配算法的精度。

[1] Zanfir A, Sminchisescu C. Deep Learning of Graph Matching. Computer Vision and Pattern Recognition, 2018.

[2] Yan J, Cho M, Zha H, et al. Multi-Graph Matching via Affinity Optimization with Graduated Consistency Regularization. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2015, 38(6):1228-1242.

[3] Yan J, Yin X, Lin W, et al. A short survey of recent advances in graph matching. International Conference on Multimedia Retrieval.

[4] Simonyan K, Zisserman A. Very Deep Convolutional Networks for Large-Scale Image Recognition. International Conference on Representation Learning, 2015.

[5] Leordeanu M, Hebert M. A spectral technique for correspondence problems using pairwise constraints. International Conference on Computer Vision, 2005