f******n
发帖数: 208 | 1 对方ID:
convolve
Feedback (+/-/0):
+
具体交易内容:
SPG points 40K
我的评价:
smooth
交易原始贴链接: |
|
k**********g
发帖数: 989 | 2 Yes.
Fixed length, fixed interval, finite time (input N time samples, output N
frequency samples)
Time domain: (1) multiply by impulse train, (2) multiply by boxcar (finite
time)
<-->
Frequency domain: (1) convolve by impulse train (hence periodic 2*pi), (2)
convolve by sinc (hence it is allowed to interpolate the N frequency
responses by a sinc function)
有错请指正，谢
Edit: not sinc, should be "periodic sinc" (a ratio between two sinc)
http://www.mathworks.com/help/toolbox/signal/diric.html;jsessioni |
|
c********s
发帖数: 123 | 3 Fortran program is beautiful!!!
program convolve
implicit none
c-----convolve lbl transmittance data with RSS filter function
integer ik, ip, nk, np, k, p, mm, nnp, nnk
parameter(np = 512, nk = 1024, nnp=4096, nnk=8192, mm=8)
real respQM(nk), respEQ(np, nk), respW(np)
real nrespQM(nnk), nrespEQ(nnp, nnk), nrespW(nnp)
real Imdir(nnk), Imdif(nnk), Irdir(nnp), Irdif(nnp), wn(nnk)
real tmp
character*20 label
print *, 'Reading RSS instrument |
|
c******a
发帖数: 11393 | 4 1 mcheng82 1017 97.9% (8) 8 (26) 0.01 (45)
2 miragedeva 720 89.1% (20) 14 (18) 0.02 (40)
3 submit 688 99.9% (3) 30 (6) 0.04 (33)
4 alive 472 75.0% (32) 30 (6) 0.06 (23)
5 mean6 427 96.0% (13) 13 (19) 0.03 (35)
6 hli1108 350 100.0% (1) 16 (15) 0.05 (28)
7 howto1 343 99.4% (4) 4 (40) 0.01 (45)
8 ivancn 335 97.4% (9) 16 (15) 0.05 (28)
9 zhangzh1985 296 65.6% (38) 26 (1... 阅读全帖 |
|
m****i
发帖数: 106 | 5 对方ID:
convolve
Feedback (+/-/0):
+
具体交易内容:
FIA
我的评价:
very smooth |
|
v******r
发帖数: 1630 | 6
http://www.mobileye.com/technology/processing-platforms/seeq2/
Mobileye EyeQ2® Vision Processor
- Two floating point, hyper-thread 64bit RISC 34KMIPS CPUs running at 324MHZ.
- Five Vision Computing Engines (VCE)
Classifier Engine – CE
• Image scaling & pre-processing units
• Pattern classifier units
•Tracker Engine-Trk
• Image warping & motion analysis unit
Pre-process, Window – PW
• Image Convolver & image pyramid units
• Computes vertical & horizontal ... 阅读全帖 |
|
i****x
发帖数: 17565 | 7 对的 高斯模糊就是对一个高斯核函数做卷积。bokeh可以把高斯核换成均匀圆分布核函
数，但计算量就大太多了。维基：
Bokeh can be simulated by convolving the image with a kernel that
corresponds to the image of an out-of-focus point source taken with a real
camera. Unlike conventional convolution, this convolution has a kernel that
depends on the distance of each image point and – at least in principle –
has to include image points that are occluded by objects in the foreground.[
20] Also, bokeh is not just any blur. To a first approximation, defocus bl... 阅读全帖 |
|
d*i
发帖数: 9453 | 8 ☆─────────────────────────────────────☆
zhxfifa (硕猫) 于 (Wed Feb 27 15:28:45 2008) 提到:
最近觉得手指关节有点疼，不知道是不是伤着了。还是动作不正确。需不需要搞个手套？
☆─────────────────────────────────────☆
wuhuarou (五花肉) 于 (Wed Feb 27 16:06:51 2008) 提到:
没因为没戴手套手指疼.
手套可以防滑脱,同时减轻老茧
套？
☆─────────────────────────────────────☆
convolve (康夫) 于 (Wed Feb 27 16:40:38 2008) 提到:
yes, it is very heloful.
☆─────────────────────────────────────☆
hawking (圆圆的就是一个大肉球) 于 (Wed Feb 27 16:44:00 2008) 提到:
我觉得推的动作不起老茧。拉得动作起茧
☆───────── |
|
a********o
发帖数: 585 | 9 The concert hall was rioting, standing ovations mixed
with shoutings and whisltes. Three encores. Savall is
a modest yet confident man. He knows what he is after.
Music is the God.
Folias, the eternal variations of the human soul, I
can always recognize you in your first four notes, no
matter how much you transfigure! To me, you are the
Spanish Bach's Chaconne, endless, convolving, yet
resounding always on the same theme... Ah, the
manyfolds of the soul, in which alternate the eternal
struggles |
|
b*******t
发帖数: 33714 | 10 基本的想法是很老的东西了
问题是psf算起来没那么容易，另外deconvolution当中涉及原片噪声（你拍到的是
object [convolve] psf + noise），没有那么好estimate |
|
b***n
发帖数: 53 | 11 请诸位继续补充, 填补各项空白
86:zhw
88:rememberance
90:rockey, mouseking
91:caterpillar&aquila,river,aladdin
92:fabian,away,pda,tangdynasty,feelingfool
93:agang,mmily,xlshen
94:barber,irix,brain,lbbg,Sakuragi,shakuras,zzzzh,lbbg
95:candid,egret,zeropoint,hunterkiller,gen,anti
96:AOL,torlim,gardenia,sturtle
97:dreamfly,tingsheng,convolve,Applecore,Aleo,stonefish,soso,cqianqian,chester
surely,whoami,whyflyaway,boung,Feily
98:qoqo,sc,bluechacha
99:arrowrooot,lakewater |
|
r****y
发帖数: 57 | 12
86: zhw, rememberance
89: yangsansan
90: mouseking, rockey, scorpio
91: caterpillar, aquila, river, aladdin
92: fabian, away, pda, tangdynasty, feelingfool, rivulet
93: agang, mmily, xlshen, wgy, journals, Robin, Fancier
94: barber,irix, brain, lbbg, Sakuragi, shakuras, zzzzh, lbbg
cchuang, Teddy, suckey, gangwu, busy, lix
95: candid, egret, zeropoint, hunterkiller, gen, anti, kentjoy
dinier
96: AOL, torlim, gardenia, sturtle
97: dreamfly, tingsheng, convolve, Applecore, Aleo, stonefish |
|
d*l
发帖数: 400 | 13
的s
The problem description says you can use any method you like. To achieve
better smoothness, I guess the medthod of "linear" should be better than
"nearest". You can also try use a gaussian function to convolve with the
rotated image.
As for how to choose the neighborhood region, I guess it is up to you. I would
make it as small as possible. Because otherwise the image will be blurred.
怎
You don't have to use any special extrapolation method. See my answer to the
following question.
。
This cl |
|
k**********g
发帖数: 989 | 14 fourier transform? (convolution to multiplication)
let the set of integers be x1, x2, x3, ...
signal representation f(x) = delta(x-x1) + delta(x-x2) + delta(x-x3) + ...
a signal that gives all possible values of xi + xj is given by
f(x) convolve f(x)
can be computed in finite number of coefficients in fourier transform.
題目的對象是誰？計算機本科？理工？數學系？ |
|
d*****u
发帖数: 17243 | 15 我不是做图像的
看该领域内好多所谓卷积其实就是把一个mask矩阵跟图像各部分逐次做乘法然后加起来
也就是correlation而不是convolution
但是他们也都说是卷积 |
|
l********a
发帖数: 1154 | 16 图像处理的卷积就是用卷积核(权矩阵)在图像上滑动,用当前窗口下的图像数据与权矩
阵做算术加权求和的结果作为结果图像的数据,这个过程有时也叫加窗,类似加权的移动
平均(Moving Average).数学上，移动平均可视为一种卷积。 |
|
|
d*****u
发帖数: 17243 | 18 一般的卷积，其中一个函数必须要反转一下
不管离散还是连续都是这样
但是图像处理里常说的卷积没有反转
楼上说这也可以看成是广义的卷积，也许是如此
不过跟一般的定义还是有差别 |
|
h*****n
发帖数: 188 | 19 这些mask大部分都是对称的，高通/低通/平均/高斯等等，所以反转无所谓
对于少数的不对称的mask，比如梯度.. 它的物理意义应该并不是卷积，只是简单的左
右相加减而已。 所以也不需要反转。 |
|
d*****u
发帖数: 17243 | 20 这个倒是
不过很多小波变换的kernel不是对称的，或者只在某一方向对称
所以一说是卷积就容易引起误会 |
|
k**********g
发帖数: 989 | 21 图像处理还是培养了不少差不多先生／女士滴，反正计算结果看起来近似就可以了 |
|
D***r
发帖数: 7511 | 22 感觉图像处理基本就是EE
现在跟machine learning结合得多，才稍微像CS了一些
不过又有人说ML不算CS |
|
w***g
发帖数: 5958 | 23 比如某层convolution, filter size是5, step size是2, 那么这层output的每个
pixel的receptive field就是7. 然后就一层层叠加。
我有个自动算receptive field的程序
https://github.com/aaalgo/tfgraph
你的training example是66, 这个一般会小于receptive field。所以你的model
train的时候padding就会生效，会pad进去synthatic的东西，倒是不一定是白边。
如果你的每个training example都需要padding, 你的model就会把padding作为
数据的特征学进去。
testing时如果图片大于receptive field，那么convolve到中间的时候就不需要
padding，这时候model看不到padding反而就觉得不正常了。
con
， |
|
m********5
发帖数: 17667 | 24 Normally 20*300 and feed to 1D conv. This means your data point is in 300D
space. And you will process the text word by word.
If you do concatenate to a 6000D vector, you will treat the text similar as
20-gram and arbitrarily define your 20-gram features as this 6000D vector,
which has been seen in many literatures. You can also do other merging
method like averaging, max, etc. This means each 20 words will be a single
data point in the space.
It does not make too much sense to me, if you conca... 阅读全帖 |
|
y********a
发帖数: 138 | 25 Dude, your view is way out of date
1. no difference bw industry & academia regarding job security.No money no
stability, holds everywhere. school==stability is an old imagination from
past generation
2. "少点功利心", are you kidding? Nowadays,you need to be as aggressive/功利
as in industry, if not more, to survive in academia. In bio, industry &
academia research are quite convolved, mainly differ at the end point:
product vs paper. In fact, "踏踏实实做research" is more common in industry,
because you don'... 阅读全帖 |
|
m***a
发帖数: 2 | 26 Hi,
I wonder how to compute the approximate inverse of a nonlinear operator. for
example, I have an operator, which convolve the input function with an
analytic function and then take the square of the convolution result. I
would like to find out the approximate inverse of such process.
does anyone know what book/papers describing such things?
Thanks
mlfma |
|
c*******v
发帖数: 2599 | 27 analytic function的Fourier级数和Chebyshev级数展开系数是指数下降的，
计算这种逆问题的时候只要记住不要用Taylor级数作就行了。
或者用待定系数法，或者取点值解代数方程组都可以。
Hi,
I wonder how to compute the approximate inverse of a nonlinear operator. for
example, I have an operator, which convolve the input function with an
analytic function and then take the square of the convolution result. I
would like to find out the approximate inverse of such process.
does anyone know what book/papers describing such things?
Thanks
mlfma |
|
s******k
发帖数: 6659 | 28 R的convolve默认是circular的…做时间序列被坑过。我怎么觉得还是matlab的函数靠
的住 |
|
h**t
发帖数: 1678 | 29 每个matrix都很大，要连续做，现在用convolve(). xiexei ! |
|
m*********k
发帖数: 10521 | 30 [Basketball] peopleocean Mar 18 ● 版主得奖俺来隆重恭喜！优秀版主！
10*100=1000
成功奖励 10 伪币的用户: andysun, AngelMAC, Antoine, anubiswu, AquamarinE1, baybear, Bohr, bond7, Brabit, BRavens, bruclee, Casillas, casio, chahu, circus, codit, convolve, crabghh, davidyulei, diehard, Drake, equator, esperlada, FinalFantasy, flashdog, flyer, gaons, Genghis, godorme, goubeibei, Grange, hsh, iso35, JianlianYi, Jingqg1979, jl0913, K24, kickkaka, laomaomao, ldxiek, ldxk, lg, lucky8, LutherHead, luxilon, lxinzheng, ma |
|
m*********k
发帖数: 10521 | 31 joe81
江小鱼骂得好 非诚勿扰女嘉宾自取其辱 bagua
bigbigbee
大飞访谈2: 旺夫女第二款 Chicago
bumpkin
大时代之前【3】音乐音乐 ChuanYu
博士工作的ideas是导师还是自己提出的? Dreamer
我是得了忧郁症了吗？ Dreamer
rivulet0707
被狗挠破以后，我是不是确实挺事多的 Family
dollarcoupon
一次party Food
mitchell1984
从历史发展角度说，黑人是不是比白人要聪明呢？ History
convolve
EB1b 140 approved in 5 days Immigration
hocc
还有人记得你么？ LeisureTime
iamyao
习惯了离别。 LosAngeles
funny88
[我也来减肥} loseweight
desmond2046
妈妈讲年轻时的故事 Military
mdphd
中 |
|
