
论文阅读笔记
man_world
这个作者很懒,什么都没留下…
-
原创 A guide to convolution arithmetic for deep learning
一、Discrete convolutionsA discrete convolution is a linear transformation that preserves this notion of ordering. It is sparse (only a few input units contribute to a given output unit) and reuses para2016-11-23 16:31:172673
0
-
原创 Image Restoration Using Deep Convolutional Encoder-Decoder Networks with Symmetric Skip Connections
一、本文的主要贡献1. 卷积和反卷积对称网络结构的提出 A very deep network architecture, which consists of a chain of symmetric convolutional and deconvolutional layers, for image restoration is proposed in this paper. The con2017-01-10 16:18:112875
11
-
转载 基于感知损失函数的实时风格转换和超分辨率重建
基于感知损失函数的实时风格转换和超分辨率重建文章地址:《Perceptual Losses for Real-Time Style Transfer and Super-Resolution》 arXiv.1603.08155转自:基于感知损失函数的实时风格转换和超分辨率重建 (zhwhong)Abstract 摘要:我们考虑的图像转换的问题,即将一个输入2017-01-12 15:38:4513718
0
-
原创 DCNN Cascade for Facial Point Detection Summary
一、总体思想—由粗到精的定位 利用人脸检测器(haar特征或者其它的detection方法),把人脸部位的图片(bbox)裁剪出来,送入CNN 好处:外界背景因素(eg:风景、身体、头发)的干扰减小了,有利于提高精度。 detection: 采用CNN定位出5个特征点的bbox精度要比直接采用人脸检测器的效果好。 利用level1的CNN粗定位出5个特征点的坐标2017-07-24 15:04:54450
0
-
原创 DCNN Cascade for Facial Point Detection Report
DCNN Cascade for Facial Point Detection Report1. 作者的模型自己跑出来的结果 Mean Error Level1 Level2 Level3 Left Eye 0.022130 0.015004 0.013403 Right Eye 0.022816 0.014660 0.2017-07-24 15:11:25346
0