计算机视觉Github开源论文

计算机视觉Github开源论文

计算机视觉Github开源论文

计算机视觉Github开源论文 U-GAT-IT Unsupervised Generative Attentional Networks with Adaptive Layer-Instance Normalization for Image-to-Image Translation

知乎转引的此文介绍了来自北京航空航天大学刘祥龙副教授研究团队的最新综述文章 **Binary Neural Networks: A Survey**，合作者包括中国电子科技大学的宋井宽教授和意大利特伦托大学计算机系主任 Nicu Sebe 教授。在阅读基础上，做了.md的笔记。 摘要如下： 神经网络二值化能够**最大程度地降低模型的存储占用和模型的计算量**，将神经网络中**原本 32 位浮点数参数量化至 1 位定点数**，**降低了模型部署的存储资源消耗，同时极大加速了神经网络的推断过程**。但二值化会不可避免地导致**严重的信息损失**，其**量化函数不连续性也给深度网络的优化带来了困难**。 近年来许多算法被提出致力于解决上述问题，并取得了令人满意的进展。在本文中，我们对这些方法进行了全面的总结和概括，主要分为**直接量化的朴素二值化方法**，以及使用**最小化量化误差**、**改善网络损失函数和减小梯度误差**等技术的**改进二值化方法**。 本文还调研了二值神经网络的**其他实用方面**，例如**硬件友好的设计和训练技巧**。然后，我们对**图像分类，目标检测和语义分割**等不同任务进行了**评估和讨论**。最后，本文展望了**未来研究可能面临的挑战**。

数据科学家，学生，机器学习工程师

汽车CAN网络安全威胁实例及短期对策选择 Security threats to automotive CAN networks—Practical examples and selected short-term countermeasures

人脸识别论文，手动翻译，花费两天时间希望可以帮助到大家 DeepID3: Face Recognition with Very Deep Neural Networks Yi Sun1 Ding Liang2 Xiaogang Wang3,4 Xiaoou Tang1,4

Networks constitute the backbone of complex systems, from the human brain to computer communications, transport infrastructures to online social systems, metabolic reactions to financial markets. Characterising their structure improves our understanding of the physical, biological, economic and social phenomena that shape our world. Rigorous and thorough, this textbook presents a detailed overview of the new theory and methods of network science. Covering algorithms for graph exploration, node ranking and network generation, among the others, the book allows students to experiment with network models and real-world data sets, providing them with a deep understanding of the basics of network theory and its practical applications. Systems of growing complexity are examined in detail, challenging students to increase their level of skill. An engaging pre- sentation of the important principles of network science makes this the perfect reference for researchers and undergraduate and graduate students in physics, mathematics, engineering, biology, neuroscience and social sciences.

Here you find all the data sets described and analysed in the textbook: "Complex Networks: Principles, Methods and Applications", V. Latora, V. Nicosia, G. Russo (Cambridge University Press, 2017) For each data set you find below a brief description and a list of salient properties (number of node, number of edges, etc.). All data sets All the data sets of the textbook are available for download in a single compressed file: All the data sets in the book (zip) The archive contains one folder for each dataset, The file README.txt in each folder contains some relevant information about the corresponding data set.