SoftPool:近似指数最大池的代码

上传者: 42097914 | 上传时间: 2022-05-05 12:39:31 | 文件大小: 105.42MB | 文件类型: ZIP
使用SoftPool完善激活下采样 抽象的 卷积神经网络(CNN)使用合并来减小激活图的大小。这个过程对于局部实现空间不变性和增加后续卷积的接收场至关重要。池操作应最大程度地减少激活图中的信息丢失。同时,应限制计算和内存开销。为了满足这些要求,我们提出了SoftPool:一种快速有效的方法,可以对指数加权的激活求和。与一系列其他合并方法相比,SoftPool在下采样的激活图中保留了更多信息。更精细的下采样可导致更好的分类准确性。在ImageNet1K上,对于一系列流行的CNN架构,用SoftPool替换原始的合并操作会导致精度不断提高1-2%。我们还将在视频数据集上测试SoftPool以进行动作识别。再次,仅替换池层将一致地提高准确性,而计算负载和内存仍然受到限制。这些有利的特性使SoftPool成为当前合并操作(包括最大池和平均池)的绝佳替代品。 基于图像的池化。图像在高度和宽度上均进行

文件下载

资源详情

[{"title":"( 45 个子文件 105.42MB ) SoftPool:近似指数最大池的代码","children":[{"title":"SoftPool-master","children":[{"title":"LICENSE.txt <span style='color:#111;'> 1.05KB </span>","children":null,"spread":false},{"title":"images","children":[{"title":"tower_soft.jpg <span style='color:#111;'> 205.50KB </span>","children":null,"spread":false},{"title":"cars.gif <span style='color:#111;'> 14.95MB </span>","children":null,"spread":false},{"title":"parkour_soft.gif <span style='color:#111;'> 2.72MB </span>","children":null,"spread":false},{"title":"parkour.gif <span style='color:#111;'> 12.42MB </span>","children":null,"spread":false},{"title":"otters.jpg <span style='color:#111;'> 45.75KB </span>","children":null,"spread":false},{"title":"puffin.jpg <span style='color:#111;'> 138.45KB </span>","children":null,"spread":false},{"title":"tower.jpg <span style='color:#111;'> 297.64KB </span>","children":null,"spread":false},{"title":"pass.gif <span style='color:#111;'> 19.13MB </span>","children":null,"spread":false},{"title":"tram.jpg <span style='color:#111;'> 81.88KB </span>","children":null,"spread":false},{"title":"tennis_ball_soft.jpg <span style='color:#111;'> 97.92KB </span>","children":null,"spread":false},{"title":"buildings_soft.jpg <span style='color:#111;'> 476.61KB </span>","children":null,"spread":false},{"title":"pass_soft.gif <span style='color:#111;'> 3.52MB </span>","children":null,"spread":false},{"title":"basketball.gif <span style='color:#111;'> 10.90MB </span>","children":null,"spread":false},{"title":"pizza_toss_soft.gif <span style='color:#111;'> 2.09MB </span>","children":null,"spread":false},{"title":"tram_soft.jpg <span style='color:#111;'> 51.94KB </span>","children":null,"spread":false},{"title":"bowling.gif <span style='color:#111;'> 17.72MB </span>","children":null,"spread":false},{"title":"basketball_soft.gif <span style='color:#111;'> 2.25MB </span>","children":null,"spread":false},{"title":"pizza_toss.gif <span style='color:#111;'> 12.09MB </span>","children":null,"spread":false},{"title":"otters_soft.jpg <span style='color:#111;'> 37.66KB </span>","children":null,"spread":false},{"title":"puffin_soft.jpg <span style='color:#111;'> 104.11KB </span>","children":null,"spread":false},{"title":"tennis_ball.jpg <span style='color:#111;'> 771.10KB </span>","children":null,"spread":false},{"title":"bowling_soft.gif <span style='color:#111;'> 2.77MB </span>","children":null,"spread":false},{"title":"cars_soft.gif <span style='color:#111;'> 2.46MB </span>","children":null,"spread":false},{"title":"buildings.jpg <span style='color:#111;'> 575.44KB </span>","children":null,"spread":false}],"spread":false},{"title":"README.md <span style='color:#111;'> 5.75KB </span>","children":null,"spread":false},{"title":"pytorch","children":[{"title":"test-files","children":[{"title":"._out_1 <span style='color:#111;'> 4.00KB </span>","children":null,"spread":false},{"title":"._test.py <span style='color:#111;'> 4.00KB </span>","children":null,"spread":false},{"title":"test.py <span style='color:#111;'> 12.99KB </span>","children":null,"spread":false},{"title":"._images <span style='color:#111;'> 4.00KB </span>","children":null,"spread":false}],"spread":true},{"title":"._setup.py <span style='color:#111;'> 4.00KB </span>","children":null,"spread":false},{"title":"CUDA","children":[{"title":"softpool_cuda_kernel.cu <span style='color:#111;'> 25.95KB </span>","children":null,"spread":false},{"title":"softpool_cuda.cpp <span style='color:#111;'> 8.06KB </span>","children":null,"spread":false},{"title":"limits.cuh <span style='color:#111;'> 5.85KB </span>","children":null,"spread":false}],"spread":true},{"title":"setup.py <span style='color:#111;'> 582B </span>","children":null,"spread":false},{"title":"Makefile <span style='color:#111;'> 126B </span>","children":null,"spread":false},{"title":"SoftPool","children":[{"title":"__init__.py <span style='color:#111;'> 191B </span>","children":null,"spread":false},{"title":"idea.py <span style='color:#111;'> 11.55KB </span>","children":null,"spread":false}],"spread":true}],"spread":true},{"title":"main","children":[{"title":"train.py <span style='color:#111;'> 19.03KB </span>","children":null,"spread":false},{"title":"models","children":[{"title":"config.py <span style='color:#111;'> 2.56KB </span>","children":null,"spread":false},{"title":"densent.py <span style='color:#111;'> 12.44KB </span>","children":null,"spread":false},{"title":"__init__.py <span style='color:#111;'> 0B </span>","children":null,"spread":false},{"title":"inception.py <span style='color:#111;'> 18.50KB </span>","children":null,"spread":false},{"title":"resnet.py <span style='color:#111;'> 16.56KB </span>","children":null,"spread":false}],"spread":true},{"title":"._train.py <span style='color:#111;'> 4.00KB </span>","children":null,"spread":false}],"spread":true}],"spread":true}],"spread":true}]

评论信息

免责申明

【只为小站】的资源来自网友分享,仅供学习研究,请务必在下载后24小时内给予删除,不得用于其他任何用途,否则后果自负。基于互联网的特殊性,【只为小站】 无法对用户传输的作品、信息、内容的权属或合法性、合规性、真实性、科学性、完整权、有效性等进行实质审查;无论 【只为小站】 经营者是否已进行审查,用户均应自行承担因其传输的作品、信息、内容而可能或已经产生的侵权或权属纠纷等法律责任。
本站所有资源不代表本站的观点或立场,基于网友分享,根据中国法律《信息网络传播权保护条例》第二十二条之规定,若资源存在侵权或相关问题请联系本站客服人员,zhiweidada#qq.com,请把#换成@,本站将给予最大的支持与配合,做到及时反馈和处理。关于更多版权及免责申明参见 版权及免责申明