<?xml version="1.0" encoding="utf-8" ?><rss version="2.0"><channel><title><![CDATA[AI小作坊 的博客]]></title><description><![CDATA[大道至简，天人合一]]></description><link>https://blog.csdn.net/zhangjunhit</link><language>zh-cn</language><generator>https://blog.csdn.net/</generator><copyright><![CDATA[Copyright &copy; zhangjunhit]]></copyright><item><title><![CDATA[Fundamental concepts about Optics]]></title><link>https://blog.csdn.net/zhangjunhit/article/details/132520581</link><guid>https://blog.csdn.net/zhangjunhit/article/details/132520581</guid><author>zhangjunhit</author><pubDate>Mon, 28 Aug 2023 09:27:02 +0800</pubDate><description><![CDATA[Book: Optics F2f  From Fourier to FresnelA particularly useful solution of the wave equation is a wave with a particular wavelength λ. This is known as the harmonic wave solution and corresponds to the case of monochromatic lightA phasor is a unit vector i]]></description><category></category></item><item><title><![CDATA[Fundamental concepts about Lithography]]></title><link>https://blog.csdn.net/zhangjunhit/article/details/132507683</link><guid>https://blog.csdn.net/zhangjunhit/article/details/132507683</guid><author>zhangjunhit</author><pubDate>Sat, 26 Aug 2023 16:32:45 +0800</pubDate><description><![CDATA[如果我们希望将 aperture 的pattern 在 image plane(reisit) 上成像，我们可以用 Collimating Lens 光照 pattern, 再用 Focusing Lens 收集光成像。Fraunhofer diffraction - far field.  弗朗霍菲衍射。Fresnel diffraction - near field.    菲涅耳衍射。]]></description><category></category></item><item><title><![CDATA[basics of level set]]></title><link>https://blog.csdn.net/zhangjunhit/article/details/127706372</link><guid>https://blog.csdn.net/zhangjunhit/article/details/127706372</guid><author>zhangjunhit</author><pubDate>Sat, 05 Nov 2022 17:43:26 +0800</pubDate><description><![CDATA[level set basics]]></description><category></category></item><item><title><![CDATA[Image Segmentation Based on Active Contours without Edges]]></title><link>https://blog.csdn.net/zhangjunhit/article/details/127597424</link><guid>https://blog.csdn.net/zhangjunhit/article/details/127597424</guid><author>zhangjunhit</author><pubDate>Sun, 30 Oct 2022 11:42:31 +0800</pubDate><description><![CDATA[Nonlinear Anisotropic Diffusion




Active Contours without Edges]]></description><category></category></item><item><title><![CDATA[Segmentation under geometrical conditions using geodesic active contours and interpolation]]></title><link>https://blog.csdn.net/zhangjunhit/article/details/127589870</link><guid>https://blog.csdn.net/zhangjunhit/article/details/127589870</guid><author>zhangjunhit</author><pubDate>Sat, 29 Oct 2022 17:49:17 +0800</pubDate><description><![CDATA[Segmentation under geometrical conditions using geodesic active contours and interpolation using level set methods

The level set approach
Using Euler–Lagrange theorem

We now propose an alternative proof (using Gâteaux derivative) of the evolution equatio]]></description><category></category></item><item><title><![CDATA[Geodesic Active Contours]]></title><link>https://blog.csdn.net/zhangjunhit/article/details/127588520</link><guid>https://blog.csdn.net/zhangjunhit/article/details/127588520</guid><author>zhangjunhit</author><pubDate>Sat, 29 Oct 2022 17:41:53 +0800</pubDate><description><![CDATA[2.1  Energy Based Active Contours
The classical snakes approach (Kass et al., 1988) associates the curve C with an energy given by
Therefore, curve smoothing will be obtained also with β = 0, having only the first regularization term

curve evolution equat]]></description><category></category></item><item><title><![CDATA[Lithography tutor]]></title><link>https://blog.csdn.net/zhangjunhit/article/details/127544599</link><guid>https://blog.csdn.net/zhangjunhit/article/details/127544599</guid><author>zhangjunhit</author><pubDate>Thu, 27 Oct 2022 13:13:41 +0800</pubDate><description><![CDATA[So how does this optical path difference affect the formation of an image? For light, the path length traveled is equivalent to a change in phase. Thus, the OPD can be expressed as a phase error,Our interpretation of defocus is that it causes a phase error]]></description><category></category></item><item><title><![CDATA[fundamental of Level set method]]></title><link>https://blog.csdn.net/zhangjunhit/article/details/127335742</link><guid>https://blog.csdn.net/zhangjunhit/article/details/127335742</guid><author>zhangjunhit</author><pubDate>Sat, 15 Oct 2022 15:28:40 +0800</pubDate><description><![CDATA[Ref:  Level Set Methods and Dynamic Implicit Surfaces]]></description><category></category></item><item><title><![CDATA[Anisotropic Diffusion in Image Processing 2]]></title><link>https://blog.csdn.net/zhangjunhit/article/details/127173372</link><guid>https://blog.csdn.net/zhangjunhit/article/details/127173372</guid><author>zhangjunhit</author><pubDate>Wed, 05 Oct 2022 15:55:42 +0800</pubDate><description><![CDATA[Parameter Adaptation for Nonlinear Diffusion in Image Processing

discrete approximation]]></description><category></category></item><item><title><![CDATA[Anisotropic Diffusion in Image Processing (1)]]></title><link>https://blog.csdn.net/zhangjunhit/article/details/127173211</link><guid>https://blog.csdn.net/zhangjunhit/article/details/127173211</guid><author>zhangjunhit</author><pubDate>Wed, 05 Oct 2022 15:44:46 +0800</pubDate><description><![CDATA[Adaptive Smoothing- A General Tool for Early Visionfor linear heat diffusion equation:
for anisotropic diffusion:
for 1D adaptive smoothing iteration equation:]]></description><category></category></item><item><title><![CDATA[MKL sparse QR solver for least square]]></title><link>https://blog.csdn.net/zhangjunhit/article/details/123567308</link><guid>https://blog.csdn.net/zhangjunhit/article/details/123567308</guid><author>zhangjunhit</author><pubDate>Fri, 18 Mar 2022 10:02:24 +0800</pubDate><description><![CDATA[COO to CSR format
#include &lt;vector&gt;
#include &lt;iostream&gt;
#include &lt;mkl.h&gt;
#ifdef __linux__
#include &lt;stdlib.h&gt; //for aligned alloc!
#include &lt;cstring&gt; //believe it or not for memcpy!!
#endif
#include "mkl_sparse_qr.h"

// -----]]></description><category></category></item><item><title><![CDATA[1D and 2D Gaussian Derivatives]]></title><link>https://blog.csdn.net/zhangjunhit/article/details/121261169</link><guid>https://blog.csdn.net/zhangjunhit/article/details/121261169</guid><author>zhangjunhit</author><pubDate>Thu, 11 Nov 2021 09:19:42 +0800</pubDate><description><![CDATA[http://campar.in.tum.de/Chair/HaukeHeibelGaussianDerivatives


The Two-Dimensional Case
Base Function (0th order)





Computes discrete 1D gaussian functions
function [ gaussian ] = gaussian( x, sigma, order, normalize )
        if isempty(normalize)
    ]]></description><category></category></item><item><title><![CDATA[MobileNet  Unet]]></title><link>https://blog.csdn.net/zhangjunhit/article/details/116980377</link><guid>https://blog.csdn.net/zhangjunhit/article/details/116980377</guid><author>zhangjunhit</author><pubDate>Tue, 18 May 2021 11:09:00 +0800</pubDate><description><![CDATA[https://github.com/bubbliiiing/Semantic-Segmentation
https://github.com/BBuf/Keras-Semantic-Segmentation
https://github.com/bubbliiiing/Semantic-Segmentation

]]></description><category></category></item><item><title><![CDATA[U-NET 图像预处理]]></title><link>https://blog.csdn.net/zhangjunhit/article/details/116479223</link><guid>https://blog.csdn.net/zhangjunhit/article/details/116479223</guid><author>zhangjunhit</author><pubDate>Fri, 07 May 2021 09:45:36 +0800</pubDate><description><![CDATA[首先将图像格式及大小、类型、名称   做出调整
这里将 bmp 转为 png  大小统一为 500*500， 按照数字序号命名
bmp_png.py

from PIL import Image
import glob
import os

out_dir = 'D:/图像数据/橙子/TestIMG/'
cnt = 501
for img in glob.glob('D:/图像数据/橙子/测试图像/*.bmp'):
    Image.open(img).resize((500,500)).save(os.]]></description><category></category></item><item><title><![CDATA[一致性直线提取--Coherent Line Drawing]]></title><link>https://blog.csdn.net/zhangjunhit/article/details/114829576</link><guid>https://blog.csdn.net/zhangjunhit/article/details/114829576</guid><author>zhangjunhit</author><pubDate>Mon, 15 Mar 2021 14:44:07 +0800</pubDate><description><![CDATA[Coherent Line Drawing
Proc. NPAR 2007
https://github.com/uva-graphics/coherent_line_drawing
https://github.com/SSARCandy/Coherent-Line-Drawing
https://ssarcandy.tw/2017/06/26/Coherent-Line-Drawing/
所谓的 Line drawing 就是直线素描，在这里的意境就是：输入一幅图像，输出一副直线艺术风格画

本文主要是]]></description><category></category></item><item><title><![CDATA[Kullback-Leibler Divergence]]></title><link>https://blog.csdn.net/zhangjunhit/article/details/112357819</link><guid>https://blog.csdn.net/zhangjunhit/article/details/112357819</guid><author>zhangjunhit</author><pubDate>Fri, 08 Jan 2021 15:01:53 +0800</pubDate><description><![CDATA[http://alpopkes.com/files/kl_divergence.pdf
Kullback-Leibler 散度
定义：  Kullback-Leibler 散度用于度量两个分布的相似性（或差异）。
对于两个离散概率分布 P 和 Q ，在一个点集合 X 上 Kullback-Leibler 散度定义如下：
DKL(P∣∣Q)=∑x∈XP(x)log(P(x)Q(x)) D_{KL}(P||Q)=\sum_{x\in X}^{}P(x)log(\frac{P(x)}{Q(x)} )  DKL​(]]></description><category></category></item><item><title><![CDATA[Bias Variance Tradeoff – Clearly Explained]]></title><link>https://blog.csdn.net/zhangjunhit/article/details/112347739</link><guid>https://blog.csdn.net/zhangjunhit/article/details/112347739</guid><author>zhangjunhit</author><pubDate>Fri, 08 Jan 2021 11:10:30 +0800</pubDate><description><![CDATA[Bias Variance Tradeoff is a design consideration when training the machine learning model. Certain algorithms inherently have a high bias and low variance and vice-versa. In this one, the concept of bias-variance tradeoff is clearly explained so you make a]]></description><category></category></item><item><title><![CDATA[What is Mahalanobis distance?   马氏距离]]></title><link>https://blog.csdn.net/zhangjunhit/article/details/112346505</link><guid>https://blog.csdn.net/zhangjunhit/article/details/112346505</guid><author>zhangjunhit</author><pubDate>Fri, 08 Jan 2021 10:39:53 +0800</pubDate><description><![CDATA[https://blogs.sas.com/content/iml/2012/02/15/what-is-mahalanobis-distance.html
https://blogs.sas.com/content/iml/2012/02/08/.html
A variance-covariance matrix expresses linear relationships between variables. Given the covariances between variables, did yo]]></description><category></category></item><item><title><![CDATA[梯度下降原理解析]]></title><link>https://blog.csdn.net/zhangjunhit/article/details/107811430</link><guid>https://blog.csdn.net/zhangjunhit/article/details/107811430</guid><author>zhangjunhit</author><pubDate>Wed, 05 Aug 2020 11:44:25 +0800</pubDate><description><![CDATA[1 原理在机器学习的核心内容就是把数据喂给一个人工设计的模型，然后让模型自动的“学习”，从而优化模型自身的各种参数，最终使得在某一组参数下该模型能够最佳的匹配该学习任务。那么这个“学习”的过程就是机器学习算法的关键。梯度下降法就是实现该“学习”过程的一种最常见的方式，尤其是在深度学习(神经网络)模型中，BP反向传播方法的核心就是对每层的权重参数不断使用梯度下降来进行优化。梯度下降法(gradient descent)是一种常用的一阶(first-order)优化方法，是求解无约束优化问题最简单、最经典的方法]]></description><category></category></item><item><title><![CDATA[加法神经网络--AdderNet: DoWe Really Need Multiplications in Deep Learning?]]></title><link>https://blog.csdn.net/zhangjunhit/article/details/106687589</link><guid>https://blog.csdn.net/zhangjunhit/article/details/106687589</guid><author>zhangjunhit</author><pubDate>Thu, 11 Jun 2020 15:15:59 +0800</pubDate><description><![CDATA[AdderNet: DoWe Really Need Multiplications in Deep Learning?
CVPR2020
https://arxiv.org/abs/1912.13200
当前主流的CNN网络使用了大量的乘法运算来计算 输入特征层和卷积滤波器的相似性（cross-correlation），由于乘法运算耗时明显大于加法运算耗时，所有本文提出一个加法神经网络，使用 l1 范数来计算 输入特征层和卷积滤波器的相似性。这样在计算滤波器的输出响应时基本不用乘法运算。针对该加法神经网络]]></description><category></category></item></channel></rss>