知识总结

1、cudnn.enabled = True和cudnn.benchmark = True作用：提高运行效率；

https://blog.csdn.net/qq_27061325/article/details/87541167

2、pytorch的loss：

https://blog.csdn.net/ls83776736/article/details/80979470

https://www.cnblogs.com/o-v-o/p/9975364.html

https://blog.csdn.net/zhangxb35/article/details/72464152

https://www.jianshu.com/p/a6131515ee1d

https://www.cnblogs.com/ranjiewen/p/10059490.html

https://blog.csdn.net/f156207495/article/details/88874911

https://blog.csdn.net/qq_22210253/article/details/85222093

https://blog.csdn.net/zziahgf/article/details/80196376

bce loss == nan：https://discuss.pytorch.org/t/bce-loss-returning-nan/2524/7

3、小人脸检测

https://zhuanlan.zhihu.com/p/48156904

https://blog.csdn.net/zhangjunhit/article/details/54601101

https://blog.csdn.net/marleylee/article/details/81177293

4、sqeence/rollaxis/transpose

https://blog.csdn.net/flysky_jay/article/details/81607289

https://blog.csdn.net/u014767662/article/details/84722582

4、数据归一化/增强

https://blog.csdn.net/z_feng12489/article/details/89205558

https://blog.csdn.net/ys1305/article/details/94332240

https://blog.csdn.net/tfygg/article/details/73251721

https://blog.csdn.net/weixin_43436824/article/details/91869211

https://www.jianshu.com/p/5adc0fd50071

5、目标遮挡问题

https://blog.csdn.net/Gentleman_Qin/article/details/84448155

https://blog.csdn.net/tianlock/article/details/88902977

6、cumsum

https://blog.csdn.net/yuansuo0516/article/details/78331568/

7、预训练模型及修改

https://blog.csdn.net/u011268787/article/details/80170482

https://blog.csdn.net/qq_34108714/article/details/90400230

8、pytorch中文文档：

https://pytorch-cn.readthedocs.io/zh/latest/torchvision/torchvision-models/

https://zhuanlan.zhihu.com/p/41423739

https://www.jianshu.com/p/45a26d278473

9、instraGAN

https://www.jianshu.com/p/923da9655d9d

10、nn.ConvTranspose2d的参数output_padding的作用

https://www.cnblogs.com/wanghui-garcia/p/10791778.html

11、医学图像分割-基于cGAN的X射线乳房肿块分割和形状分类

https://mbd.baidu.com/newspage/data/landingsuper?context=%7B%22nid%22%3A%22news_9234972258094308615%22%7D&n_type=1&p_from=3

12、迁移学习中为不同的网络部分设置不同的学习率

https://www.cnblogs.com/shiyublog/p/10886426.html

https://blog.csdn.net/guls999/article/details/84971882

https://blog.csdn.net/wangbin12122224/article/details/79949824

https://blog.csdn.net/qq_42110481/article/details/81025575

https://blog.csdn.net/qq_34914551/article/details/87699317

https://blog.csdn.net/qq_27261889/article/details/90237008------Pytorch的add_param_group使用说明

13、PyTorch学习之十种优化函数

https://blog.csdn.net/shanglianlm/article/details/85019633

https://blog.csdn.net/u013382233/article/details/85948695------PyTorch的自适应池化Adaptive Pooling

14、GAN的优化训练

https://blog.csdn.net/duyue3052/article/details/82862841

https://blog.csdn.net/lynlindasy/article/details/87902435

https://blog.csdn.net/AMDS123/article/details/78388247

https://github.com/soumith/ganhacks#authors

https://blog.csdn.net/sparkkkk/article/details/72598041

http://www.sohu.com/a/125598465_473283

15

图像检索：Cross-domain Image Retrieval with a Dual Attribute-aware Ranking Network

https://blog.csdn.net/Julialove102123/article/details/80838431

16、Pytorch入门学习（九）---detach()的作用（从GAN代码分析)

https://blog.csdn.net/qq_39709535/article/details/80804003

17、DCGAN、WGAN、WGAN-GP、LSGAN、BEGAN原理总结及对比

https://www.cnblogs.com/bonelee/p/9166122.html

18、学习笔记：图像风格迁移

https://blog.csdn.net/czp_374/article/details/81185603

https://mbd.baidu.com/newspage/data/landingsuper?context=%7B%22nid%22%3A%22news_9336962106420590015%22%7D&n_type=0&p_from=1

谁说只有VGG才能做风格迁移，ResNet也可以！答案就在对抗攻击中

19、pytorch 学习笔记之编写 C 扩展,又涨姿势了

自定义模块，并用pytorch调用

https://www.cnblogs.com/hongge66/p/6815114.html

20、从视频理解到目标检测，百度大脑斩获10项CVPR 2019竞赛冠军

https://mbd.baidu.com/newspage/data/landingsuper?context=%7B%22nid%22%3A%22news_9055847988444218288%22%7D&n_type=0&p_from=1

21、成熟的目标检测，也该自己学习数据增强策略达到SOTA了

https://mbd.baidu.com/newspage/data/landingsuper?context=%7B%22nid%22%3A%22news_9947934166526226893%22%7D&n_type=0&p_from=1

22、Pytorch之Tensor和Numpy之间的转换

https://blog.csdn.net/shey666/article/details/85337212

23、AAAI 2019：把Cross Entropy梯度分布拉‘平’，就能轻松超越Focal Loss

https://zhuanlan.zhihu.com/p/55017036

ICCV2019的阿里DR loss，解决目标检测正负样本不平衡问题

https://baijiahao.baidu.com/s?id=1642994473784990026&wfr=spider&for=pc

https://mbd.baidu.com/newspage/data/landingsuper?context=%7B%22nid%22%3A%22news_9354933606779902933%22%7D&n_type=0&p_from=1

https://blog.csdn.net/tiandiwoxin92/article/details/78311173

https://www.jianshu.com/p/756baa220882

https://blog.csdn.net/StreamRock/article/details/81096105

https://www.jianshu.com/p/29948a5c5560

https://github.com/lvpengyuan/masktextspotter.caffe2

https://github.com/HCIILAB/Scene-Text-Detection

https://www.sohu.com/a/242862623_418390

https://blog.csdn.net/wmsbeijing/article/details/79063043

http://xiaofengshi.com/2019/01/05/%E6%B7%B1%E5%BA%A6%E5%AD%A6%E4%B9%A0-OCR_Overview/

https://python.ctolib.com/xiaofengshi-chinese-ocr.html

https://github.com/ZumingHuang/awesome-ocr-resources

https://www.jianshu.com/p/003c5b6477ba

https://baike.baidu.com/item/%E8%A1%8C%E4%BA%BA%E9%87%8D%E8%AF%86%E5%88%AB/20815009?fr=aladdin

https://www.cnblogs.com/skyfsm/p/8443107.html

https://cloud.tencent.com/developer/article/1152494

http://www.sohu.com/a/231973296_418390

https://blog.csdn.net/weixin_37993251/article/details/88916913

https://blog.csdn.net/hao5335156/article/details/80607732

https://news.pconline.com.cn/1033/10334204.html

https://www.sohu.com/a/212109803_99963310

https://blog.csdn.net/Tencent_TEG/article/details/79907298

https://blog.csdn.net/Tencent_TEG/article/details/93684410

https://www.jianshu.com/p/afe3f3028f97?from=message&isappinstalled=0

https://blog.csdn.net/yH0VLDe8VG8ep9VGe/article/details/84207254

https://github.com/jwyang/faster-rcnn.pytorch

https://blog.csdn.net/imPlok/article/details/95041472

https://www.jianshu.com/p/7a2b227896a8

https://blog.csdn.net/javastart/article/details/96584557

https://cloud.tencent.com/developer/news/374080

https://blog.csdn.net/weixin_38708130/article/details/83789457

https://www.jianshu.com/p/14141f8b94e5

https://blog.csdn.net/CSDN_71560364126/article/details/85217989

https://cloud.tencent.com/developer/article/1486521

https://blog.csdn.net/haoji007/article/details/80361587

https://github.com/rbgirshick/py-faster-rcnn#beyond-the-demo-installation-for-training-and-testing-models

https://blog.csdn.net/weixin_43915709/article/details/89430843

http://www.mamicode.com/info-detail-2217311.html

https://blog.csdn.net/jizhidexiaoming/article/details/82502724

https://blog.csdn.net/qq_41872630/article/details/86661800

https://pytorch-cn.readthedocs.io/zh/latest/package_references/torch-nn/#_1

https://www.cnblogs.com/wanghui-garcia/p/10877700.html

https://www.jianshu.com/p/8013aa6d1f4c

https://blog.csdn.net/weixin_40100431/article/details/84349470