深度学习 资源汇总

视线估计  

1. Gaze Estimation 简介(一)   https://mp.weixin.qq.com/s/KfVQrzR23nT0wF0EPGZO8Q

2. EfficientNet  https://arxiv.org/abs/1905.11946

3. EfficientDet  https://arxiv.org/abs/1911.09070

4. Gpipe  http://papers.nips.cc/paper/8305-gpipe-efficient-training-of-giant-neural-networks-using-pipeline-parallelism.pdf

5. Spiking-YOLO  https://mp.weixin.qq.com/s/tcs2j91lNCJX-gnoJs2FlA 脉冲神经网络再目标检测的首次尝试

6. 何恺明团队最新力作RegNet：超越EfficientNet，GPU上提速5倍，这是网络设计新范式 https://mp.weixin.qq.com/s/TAOQXQ7YDVbsp_UVfvgADQ

7.NAS应用论文：

https://arxiv.org/abs/1611.01578

https://arxiv.org/abs/1806.09055

https://arxiv.org/abs/1905.11946

8.网络搜索空间：http://openaccess.thecvf.com/content_ICCV_2019/html/Radosavovic_On_Network_Design_Spaces_for_Visual_Recognition_ICCV_2019_paper.html

9.ResNetXt http://openaccess.thecvf.com/content_cvpr_2017/html/Xie_Aggregated_Residual_Transformations_CVPR_2017_paper.html

10.NAS相关网络

http://openaccess.thecvf.com/content_cvpr_2018/html/Zoph_Learning_Transferable_Architectures_CVPR_2018_paper.html

https://wvvw.aaai.org/ojs/index.php/AAAI/article/view/4405

http://openaccess.thecvf.com/content_ECCV_2018/html/Chenxi_Liu_Progressive_Neural_Architecture_ECCV_2018_paper.html

https://arxiv.org/abs/1802.03268

https://arxiv.org/abs/1806.09055

https://arxiv.org/abs/1905.11946

11.a methodology for comparing and
analyzing populations of networks sampled from a design
space.http://openaccess.thecvf.com/content_ICCV_2019/html/Radosavovic_On_Network_Design_Spaces_for_Visual_Recognition_ICCV_2019_paper.html

12.network  design  space:http://openaccess.thecvf.com/content_ICCV_2019/html/Radosavovic_On_Network_Design_Spaces_for_Visual_Recognition_ICCV_2019_paper.html

13. standard residual bottlenecks block with group convolution:http://openaccess.thecvf.com/content_cvpr_2017/html/Xie_Aggregated_Residual_Transformations_CVPR_2017_paper.html

14.超越yolov3和centernet：更快更准确的saccadenet来了：https://mp.weixin.qq.com/s/7Yjexkwv58VJdOVYoP_2tQ

15. 51.2 mAP! 商汤提出目标检测新网络TSD | CVPR2020   https://mp.weixin.qq.com/s/v_5BX1ikcOHDknfHkPAcxw

16. 一文看尽8篇目标检测最新论文（EfficientDet/EdgeNet/ASFF/RolMix/SCL/EFGRNet等）https://cloud.tencent.com/developer/article/1549566

17. 25篇最新CV领域综述来了！涵盖15个方向：目标检测/图像分割/医学影像/人脸识别等方向https://mp.weixin.qq.com/s/cCJrBvV4DC87pFCqCL_J7Q

18.小目标检测  https://cloud.tencent.com/developer/article/1436733

19.小人脸检测 https://cloud.tencent.com/developer/article/1435361

20.SENet(CVPR2018) http://openaccess.thecvf.com/content_cvpr_2018/html/Hu_Squeeze-and-Excitation_Networks_CVPR_2018_paper.html

21. PSPNet(CVPR2017) http://openaccess.thecvf.com/content_cvpr_2017/html/Zhao_Pyramid_Scene_Parsing_CVPR_2017_paper.html 

22. DeepLabV3 https://arxiv.org/abs/1706.05587

23.ShuffleNetV2(ECCV 2018)  http://openaccess.thecvf.com/content_ECCV_2018/html/Ningning_Light-weight_CNN_Architecture_ECCV_2018_paper.html 

24.Inception-v2(CVPR 2016) https://www.cv-foundation.org/openaccess/content_cvpr_2016/html/Szegedy_Rethinking_the_Inception_CVPR_2016_paper.html

25.knowledge distillation https://arxiv.org/abs/1503.02531

26.mixup training https://arxiv.org/abs/1710.09412

27. 快到没朋友的YOLOv3有了PaddlePaddle预训练模型，精度更高了https://www.csdn.net/article/a/2019-05-17/15973930

28. 137% YOLOv3加速、10倍搜索性能提升！这样的惊喜，最新版paddleslim有10个 https://m.sohu.com/sa/379076353_464065

29. Label Smoothing（Hinton老带的谷歌大脑团队发的论文 nips 2019） http://papers.nips.cc/paper/8717-when-does-label-smoothing-help

30.Dynamic R-CNN：通过动态训练实现高质量的目标检测,COCO上50.1 AP  https://wx.zsxq.com/mweb/views/topicdetail/topicdetail.html?topic_id=241411552811551&group_id=142181451122&user_id=28514284588581

31.#CVPR2020# Xsense提出：基于毫米波雷达的目标检测新网络 https://wx.zsxq.com/mweb/views/topicdetail/topicdetail.html?topic_id=421211481228158&group_id=142181451122&user_id=28514284588581

32.DMNet：航空影像中的密度图引导的目标检测网络  https://wx.zsxq.com/mweb/views/topicdetail/topicdetail.html?topic_id=241411588412111&group_id=142181451122&user_id=28514284588581

33. 提升小目标检测的思路 https://mp.weixin.qq.com/s/aV_nmsnH7LyHkFki3Zbq2g

34. 打遍天下无敌手，却说它只是个baseline！多目标跟踪FairMOT的烦恼  https://mp.weixin.qq.com/s/7IzqglD0e89Y2Oai9Jv2XQ

35.数据增强（小目标检测中的数据扩增）方面论文推荐：Augmentation for small object detection

36. Augmentation for small object detection对应的相关博客链接

https://blog.csdn.net/abrams90/article/details/89371797

https://www.cnblogs.com/xuanyuyt/p/11328548.html

37. DR loss，解决样本不均衡问题，出自阿里巴巴，刚开源，直接替代focal loss，map提升2.6个点 https://wx.zsxq.com/mweb/views/topicdetail/topicdetail.html?topic_id=548488254422444&group_id=142181451122&user_id=28514284588581

38. 闲话模型压缩之网络剪枝（Network pruning）综述 https://mp.weixin.qq.com/s/dpwS96kEqcaHWiDi2g1d2w

39.SiamAttn:目标跟踪新网络，表现SOTA，性能优于DIMP、SiamRPN、SiamDW和ATOM等，来自码隆科技https://wx.zsxq.com/mweb/views/topicdetail/topicdetail.html?topic_id=421211425188488&group_id=142181451122&user_id=28514284588581

40.模型量化：https://mp.weixin.qq.com/s/yC2Jb4feobD1MttblHw_xg

41.resnet最强改进版来了！ResNeSt：split-attention networks https://mp.weixin.qq.com/s/ENQualf9sf8y5hAMAqlf0Q

42.pytorch深度学习模型压缩开源库（含量化、剪枝、轻量化结构、bn融合）https://mp.weixin.qq.com/s/1jog9OY5EVzp1mz-ayDL-w

43.深度学习中的目标追踪概述（VOT in DeepLearning） https://www.jianshu.com/p/7bcb18084966

44.这三篇论文项目开源了！何恺明等人的MoCov2，小目标检测TinyPerson和实例分割SOLOhttps://mp.weixin.qq.com/s?__biz=MzUxNjcxMjQxNg==&mid=2247498196&idx=2&sn=422f64b69ba9e411feed3b3e99e9e49d&chksm=f9a18b5bced6024dc6efe7d69952eb27d821f47ccb0a5e9df5ef33965f2e5d63cdc49d68ee7f&mpshare=1&scene=1&srcid=&sharer_sharetime=1587352282043&sharer_shareid=cbbef2c605fddacb2d7920dbc8c32142&key=cf6ab1fd6f95a27060c9fed653b2324841905af79b3dd72fe980b5a288ed5ac21d6f6095819b41310e9101ea64457c9473fb58bbe7e3b22556a8888320cd4f0781b32efe77e4595bdc312ba1ffc10cf6&ascene=1&uin=Njk1MTg4ODgy&devicetype=Windows+10&version=62080079&lang=zh_CN&exportkey=AcI%2BdOFaUvNKQIZCnt17ai0%3D&pass_ticket=YC9SXj5ERZviWunEg866Pckknakke7yeQl1hsPOOAZbQPSf66jsWjuZj5mCF69cO

45.拯救小目标检测！Tiny Person数据集和SM尺度匹配小目标检测新方法https://mp.weixin.qq.com/s?__biz=MzUxNjcxMjQxNg==&mid=2247495001&idx=3&sn=f301890b89165916368251a20257edbd&chksm=f9a19fd6ced616c0abaaf0d7e4c9725f630e000d2bc39b76a011c28bf3d4c4ef1460dba8ab35&scene=21#wechat_redirect

46.LeCun力荐，PyTorch官方权威教程书来了，意外的通俗易懂 https://zhuanlan.zhihu.com/p/93131985?utm_source=wechat_session&utm_medium=social&utm_oi=767364532031983616 

47. 商汤研究院高级研究院51页PPT详解模型量化[附PPT下载]https://mp.weixin.qq.com/s/_v8XsQ5MMmbPcQwfkmQODw

48. YOLOv4来了，coco ap 43.5%，速度65fps， https://arxiv.org/pdf/2004.10934.pdf

49.拯救小目标检测！Stitcher:简单粗暴，高效涨点  https://mp.weixin.qq.com/s/Au7M8pDj5293bHe5smY7UQ

50.YOLO来了！你只需训练一次，谷歌大脑提出调参新trick https://mp.weixin.qq.com/s/jpO_3NHdeewrOpzDnX3v6w

51. 一种新的模型量化压缩技术，仅量化权重子集而非整个网络的压缩方案，可对模型进行极致压缩，让神经网络在不降低性能的前提下，内存占用减少90%以上  https://mp.weixin.qq.com/s/sypJWeclckv1aQaPL7Rd6Q

52.IoU、GIoU、DIoU、CIoU损失函数的那点事儿https://mp.weixin.qq.com/s/Sg9nJkylnaKEqsNZTnlM5g

53.基于YOLOv3实现轻量级人脸检测（含人脸关键点检测）https://mp.weixin.qq.com/s/1ArEl49wYyVbe2uQaSSeeg

54.商汤提出应对尺度变化的检测新算法 https://mp.weixin.qq.com/s/BPZku2DdNek7brlRO6djog

55.EQL：解决长尾下的目标识别问题 https://mp.weixin.qq.com/s/M0qqbUAymnnY9-uW3ZLFsA

56. 54.7 AP! 最强的目标检测网络：DetectorRS(已开源)https://mp.weixin.qq.com/s/CSUMpTb6w8ruaHEUThL-Tg

57.目标检测领域论文

58.大神没交棒，但YOLOv5来了：https://mp.weixin.qq.com/s/g50Lcj1mX-xegITlzhaCuQ 

59.CVPR 2020 Oral | DMCP:可微分的深度模型剪枝算法解读https://mp.weixin.qq.com/s/j_Fe0Zc3vmyN3mIZ66G1XA

60.CVPR 2020主会论文 http://openaccess.thecvf.com/CVPR2020.py

61.CVPR 2020workshop http://openaccess.thecvf.com/CVPR2020_workshops/menu.py
62.YOLOv5来了！基于PyTorch，体积比YOLOv4小巧90%，速度却超2倍https://mp.weixin.qq.com/s/QGPZQN4-nAMONtIrWQzBfQ

63. PPYOLO：2020不容错过的目标检测调参Trickshttps://mp.weixin.qq.com/s/pHOFqFihkkRVTYbkSTlG4w