https://github.com/sddai/repo/blob/master/copy_file_from_list.py
https://github.com/sddai/repo/blob/master/copy_files.py
https://github.com/sddai/repo/blob/master/data_rename.py
https://github.com/sddai/repo/blob/master/keras_to_pb.py
https://github.com/sddai/repo/blob/master/make_train_test_list.py
https://github.com/sddai/repo/blob/master/remove_blank_from_filename.py
https://github.com/sddai/repo/blob/master/resize_pic.py
https://github.com/sddai/repo/blob/master/shuffle_train_test_split.py
https://github.com/sddai/repo/blob/master/ssd_pascal_%E8%A7%A3%E8%AF%BB%E7%89%88.py
https://github.com/sddai/repo/blob/master/train_test_split.py
https://github.com/sddai/repo/blob/master/train_test_val_split_func.py
https://github.com/sddai/CV-Tools-and-DL-Resources-by-Sida-Dai/blob/master/create_datasets.py
https://github.com/sddai/CV-Tools-and-DL-Resources-by-Sida-Dai/blob/master/rename_files.py
https://github.com/sddai/CV-Tools-and-DL-Resources-by-Sida-Dai/blob/master/split_trainval.py
http://parnec.nuaa.edu.cn/huangsj/alipy/index.html
https://blog.csdn.net/xunan003/article/list/3
https://www.jianshu.com/p/a76c18a3c6d5
https://www.cnblogs.com/denny402/default.html?page=8
http://ilearning.huawei.com/edx/next/#/
EI、SCI https://blog.csdn.net/xunan003/article/details/80425625
https://www.bilibili.com/video/av35817925/?spm_id_from=333.788.videocard.1
https://www.bilibili.com/video/av35939892/?spm_id_from=333.788.videocard.6
https://www.bilibili.com/video/av11640962/
https://blog.csdn.net/cust_hf/article/category/345853
caffe视频教程：
https://www.bilibili.com/video/av28240307/
https://www.bilibili.com/video/av5912510/
https://www.bilibili.com/video/av37229163/
caffe博客：
https://www.cnblogs.com/denny402/
http://www.cnblogs.com/denny402/tag/caffe/
caffe的ROI源码实现（如何实现未定义的层）
https://blog.csdn.net/jiongnima/article/details/80016683
https://qixinbo.info/2018/07/24/vae/
https://kexue.fm/archives/5253
https://www.imagemagick.org/script/download.php
https://arxiv.org/pdf/1709.00643.pdf
Unet++
CAN32+resnet(dncnn)
Unet++ Unet++(deep supervise) CAN32+adaptive normalization CAN32 FFDNet FFDNet+dilation FFDNet+dilation[1_2_5_9] Unet(baseline)
人脸属性数据集
http://mmlab.ie.cuhk.edu.hk/projects/CelebA.html
https://zhuanlan.zhihu.com/p/41779996
Google AI blog 论文
https://ai.google/research/pubs/pub44873
https://zhuanlan.zhihu.com/p/31865988
https://github.com/AcceptedDoge/machine-learning-yearning-cn/blob/master/MLY-zh-cn.pdf
吴恩达 https://www.bilibili.com/video/av35144344?spm_id_from=333.788.b_765f64657363.1
无痛的机器学习第一季
https://zhuanlan.zhihu.com/p/22464594
深度学习经典、前沿论文讲解（知乎专栏） https://zhuanlan.zhihu.com/liuyan0612
https://blog.csdn.net/v_JULY_v/article/details/80170182
小豆豆之人脸检测（知乎） https://zhuanlan.zhihu.com/p/50929457
https://zhuanlan.zhihu.com/c_181106394
目标检测入门： https://juejin.im/entry/5a98edc06fb9a028c9797fc9
检测算法总结： https://handong1587.github.io/deep_learning/2015/10/09/object-detection.html
https://zhuanlan.zhihu.com/p/39579528
1 Faster RCNN https://zhuanlan.zhihu.com/p/30621997
2 Yolo https://blog.csdn.net/hrsstudy/article/details/70305791
源码解析： https://blog.csdn.net/column/details/13752.html
3 SSD https://blog.csdn.net/u010167269/article/details/52563573 （写的很好）
https://zhuanlan.zhihu.com/p/29410169 （写的很好）
https://zhuanlan.zhihu.com/p/31427288
https://zhuanlan.zhihu.com/p/24954433
5 RFCN https://zhuanlan.zhihu.com/p/47579399
http://caffe.berkeleyvision.org/tutorial/solver.html
https://ethereon.github.io/netscope/#/editor
目标检测算法总结与对比【写的好】 https://www.cnblogs.com/venus024/p/5590044.html
https://handong1587.github.io/deep_learning/2015/10/09/object-detection.html
超分（https://github.com/huangzehao/caffe-vdsr）
tensorflow(https://blog.csdn.net/xierhacker/article/category/6511974)
pytorch(https://blog.csdn.net/u014380165/article/details/78525273)
降噪的也有一系列代码可以参考（https://github.com/wenbihan/reproducible-image-denoising-state-of-the-art）
https://zhuanlan.zhihu.com/p/32206896
专利查询网站 https://patentscope.wipo.int/search/zh/result.jsf
caffe未定义的层： CReLU 层为PVANet中特殊的层结构，其结构如下，在Caffe中并没有标准的CReLU 层作为单独的一层。 注意：在CReLU 最早提出的论文中《Understanding and Improving Convolutional Neural Networks via Concatenated Rectified Linear Units》（http://cn.arxiv.org/abs/1603.05201）， 并没有如图3-8 所述的Scale/Shift层，也即没有训练参数。不过，在NNIE mapper支 持的单独CReLU 层的实现中，是以PVANet网络中的方式来实现，Caffemodel模型文 件中需要包含Scale层对应的参数。如果要使用不同与此方式实现的CReLU 层，可以 通过多个Caffe标准层组合实现。
PassThrough层为Yolo v2中的一个自定义层
Depthwise Convolution 层为Xception 网络中的自定义层，在caffe 框架中同样也是没有 进行标准定义的。Depthwise Convolution 实现的操作为针对输入的每一个channel，单 独做KK的卷积，假设输入的channel是M1，输出结果的channel依然是M1。因 此，可以认为有M1个KK*1 的卷积kernel。
RReLU 层即Randomized Leaky ReLU，在https://arxiv.org/abs/1505.00853中有提到，并 在文章中通过MXNet框架进行实现，在caffe 框架中同样也是没有进行标准定义的。
Permute层实现的功能是将一个多维的Tensor的维度顺序进行切换，例如将一个 NCHW顺序的Tensor转换为NHWC顺序的Tensor。
PSROIPooling层的操作与ROIPooling层类似，不同之处在于不同空间维度输出的图片 特征来自不同的feature map channels，且对每个小区域进行的是Average Pooling，不同 于ROIPooling 的Max Pooling
Upsample层为Pooling层的逆操作，下图为一个示意图，其中每个Upsample层均与网 络之前一个对应大小输入、输出Pooling层一一对应，完成feature map在spatial维度 上的扩充。
人脸关键点检测
https://mp.weixin.qq.com/s/w-ow_BP8FynTlimqBBjc8A
【博客】
https://joshua19881228.github.io/
量化
https://zhuanlan.zhihu.com/p/25382177
大全
https://freshmou.github.io/2018/10/NNIE/
https://blog.csdn.net/szx940213
Pruning Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Coding. ICLR, 2016. CNNpack: Packing Convolutional Neural Networks in the Frequency Domain. NIPS, 2016. Pruning convolutional neural networks for resource efficient transfer learning. ICLR, 2017. ThiNet: A Filter Level Pruning Method for Deep Neural Network Compression. ICCV, 2017. Packing Convolutional Neural Networks in the Frequency Domain. TPAMI, 2018. Frequency-Domain Dynamic Pruning for Convolutional Neural Networks. NIPS2018.
Teacher-Student Paradigm Distilling the Knowledge in a Neural Network. NIPS workshop, 2014. FitNets: Hints for Thin Deep Nets. ICLR, 2015. Like What You Like: Knowledge Distill via Neuron Selectivity Transfer. 2017. Paying More Attention to Attention: Improving the Performance Of Convolutional Neural Networks via Attention Transfer. ICLR, 2017. Learning from Multiple Teacher Networks. ACM SIGKDD, 2018. Adversarial Learning of Portable Student Networks. AAAI, 2018. Quantization Mimic: Towards Very Tiny CNN for Object Detection. ECCV, 2018.
Decomposition of NN Layers MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Application. Google, 2017. ShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile Devices. CVPR, 2018. Shift: A Zero FLOP, Zero Parameter Alternative to Spatial Convolutions. CVPR, 2018. MobileNetV2: Inverted Residuals and Linear Bottlenecks. CVPR, 2018. ShuffleNetV2: Practical Guidelines for Efficient CNN Architecture Design. ECCV, 2018. Learning Versatile Filters for Efficient Convolutional Neural Networks. NIPS, 2018.
NAS (Neural Architecture Search) Neural Architecture Search With Reinforcement Learning. ICLR, 2017. Large-Scale Evolution of Image Classifiers. ICML, 2017. Genetic CNN. ICCV, 2017. Regularized Evolution for Image Classifier Architecture Search. Google, 2018. Towards Evolutionary Compression. ACM SIGKDD, 2018. AMC: Automated Model Compression and Acceleration with Reinforcement Learning. ECCV, 2018. Neural Architecture Optimization. NIPS, 2018. Towards Evolutionary Compression. ACM SIGKDD, 2018.
rethinking imagenet pre-training Kaiming He
行人再识别 有监督： 表征学习 [Zheng Z et al, ACM TMCCA, 2017.] 度量学习 [M. Koestinger et al, CVPR, 2012] 迁移学习： 共享嵌入空间 [Peng et al., CVPR, 2016] 借助额外信息 [Wang et al., CVPR, 2018] 用GAN去风格迁移 [Wei et al., CVPR, 2018]
PUL(ACM TMCCA，2016)
https://medium.freecodecamp.org/debugging-tensorflow-a-starter-e6668ce72617
YOLO源码解析：
https://www.cnblogs.com/zyly/p/9534063.html
TensorFlow Android: https://blog.csdn.net/LiJiancheng0614/article/details/78095521
https://ditto-cp.sourceforge.io/
http://huyanapp.com/portal.php
Python http://www.runoob.com/python/python-tutorial.html Linux http://www.runoob.com/w3cnote/linux-common-command.html
Andrew ng的经典机器学习教程 https://www.coursera.org/learn/machine-learning 斯坦福Li Feifei的CS231深度学习（计算机视觉）课 http://cs231n.stanford.edu/ 深度学习中文参考电子书 https://tigerneil.gitbooks.io/neural-networks-and-deep-learning-zh/content/
https://blog.csdn.net/fendouaini/article/details/79885721
例如 https://baijiahao.baidu.com/s?id=1609479661013467934&wfr=spider&for=pc
例如 https://www.jianshu.com/p/85364ec88136
4.6 常用训练技巧tricks https://zhuanlan.zhihu.com/p/51975781 http://lamda.nju.edu.cn/weixs/project/CNNTricks/CNNTricks.html
5 目标检测
参考 https://zhuanlan.zhihu.com/p/29102671 https://zhuanlan.zhihu.com/c_193856688 。 https://github.com/search?q=RFCN