Deconvolution/Transposed convolution
来源:互联网 发布:高中数学讲课视频软件 编辑:程序博客网 时间:2024/06/06 12:59
老早就琢磨过deconvolution的问题,不过今天又想起这个,发现理解得还是比较模糊,具体就是还是不知道caffe或者tensorflow中实现deconvolution是怎么实现的,刚才看到下面这篇文章
https://arxiv.org/ftp/arxiv/papers/1609/1609.07009.pdf
里面的两张图一下子就把我的所有疑惑解开了:
Fig.2描述的是卷积过程,caffe中的im2col就是干的这件事
Fig.3描述的是deconvolution/transposed convolution的过程,很简单的理解就是把先前卷积用的C矩阵转置了一下。
(PS:这里之所以说C矩阵是沿用了这篇文章的叫法)
嗯。 本文给出的两个链接基本把deconvolution解释清楚了,另外theano专门有一个convolution arithmetic turorial(链接在这里)也讲了这个,不过应该先前的两个链接应该就够了 :)
来自distill上的google brain的一篇文章——deconvolution有毒啊。。会造成 checkerboard pattern
https://distill.pub/2016/deconv-checkerboard/
文中提到了他们尝试解决checkerboard pattern的一个办法:
We’ve had our best results with nearest-neighbor interpolation, and had difficulty making bilinear resize work. This may simply mean that, for our models, the nearest-neighbor happened to work well with hyper-parameters optimized for deconvolution. It might also point at trickier issues with naively using bilinear interpolation, where it resists high-frequency image features too strongly. We don’t necessarily think that either approach is the final solution to upsampling, but they do fix the checkerboard artifacts.
Resize-convolution layers can be easily implemented in TensorFlow using tf.image.resize_images(). For best results, use tf.pad() before doing convolution with tf.nn.conv2d() to avoid boundary artifacts.
- Deconvolution/Transposed convolution
- Transposed Convolution, Fractionally Strided Convolution or Deconvolution
- 反卷积(Transposed Convolution, Fractionally Strided Convolution or Deconvolution)
- 转置卷积 Transposed convolution
- Convolution Networks 和Deconvolution Networks
- Convolution Networks 和Deconvolution Networks
- Convolution Networks 和Deconvolution Networks
- [Tensorflow]2.转置卷积(Transposed Convolution)
- 【深度学习】反卷积(transposed convolution)介绍
- Deconvolution
- Deconvolution
- FCN网络中使用的caffe类型层汇总--Convolution/Deconvolution/Crop/Eltwise/SoftmaxWithLoss
- Convolution
- Convolution
- Convolution
- Richardson–Lucy deconvolution
- Deconvolution Network简单理解)
- Deconvolution Network简单理解
- 图论最大网络流增广路算法详解与实现
- Java 包与包之间的访问 (类在不同包里)
- POJ 2112 Optimal Milking (二分+floyd+二分图多重匹配)
- 170816 WarGames-Natas(0-14)
- 数据结构实验之排序七:选课名单
- Deconvolution/Transposed convolution
- JNI:在JNI中调用第三方库时所需的CMakeLists 配置
- SSL P2701 交错匹配
- 实用工具vi
- Subject 与 Observable 的区别
- 查询语言模型
- JavaScript封装
- 洛谷P3781 [TJOI2010] Middle 中位数 详细题解
- 21.odoo入门——杂记