由买买提看人间百态

topics

全部话题 - 话题: decode
首页 上页 1 2 3 4 5 6 7 8 9 10 下页 末页 (共10页)
R*R
发帖数: 2661
1
来自主题: Hardware版 - 8600GT 看1080p的问题
thanks man. Yes I tried to set up h.264 through hardware decoding using mpc-
hc. When de-select dxva, mpc-hc can play my 1080p without any problem, but
CPU usage is ~50%; when select dxva, CPU usage is ~15% but the playback
stutters.
Let me try wanmei jiema when I go back home.
Meanwhile, I am looking for a more powerful GPU for hardware decoding 1080p.
Will a GT240 or a GT430 do it? which one is better among these two cards?
i***l
发帖数: 9994
2
Many thanks man! Yes it looks like intel hd 3000 should have no problem
decoding 1080p60.
there's only one problem i have now: is this ducks take off video compressed
or raw?
the video files i have is in m2ts format, do you have 1080p60 in m2ts format
and if so, would you please help test if intel hd 3000 can decode it
smoothly?
TIA, brother!
i***l
发帖数: 9994
3
thanks kinecty for testing the sample videos!
I am very impressed by the decoding ability of HD 3000! Yes i was totally
wrong at the beginning whe i thought the HD 3000 was unable to decode
1080p60.
now I have to return this h61m-p23 and get a z68 with HDMI out.
a***e
发帖数: 27968
4
来自主题: Hardware版 - 推土机 preview,悲剧了。
整个设计理念就是降低单核性能省晶体管,增加核
所以单核几乎肯定比K10慢,跟SNB比更慢
这个应该和调度关系不大
全新的乱序执行和分支预测,这个多牛有待观察
intel的code base更大,似乎分支预测总是比amd的靠谱
另外那个decoder似乎必然成为瓶颈,一个decoder喂俩核
倒是fpu部件比较有创意
cache的问题,8M的尺寸,必然L2比512K的慢很多,比较器就那德行
其实光看延迟不差,但是带宽不行要么是所谓没优化,也有可能就是位宽不行俩核抢道
比较意外的是DDR3的性能,这个东西应该对程序透明了
K10支持ddr3么?
至于内部数据,其实看定价就知道了,这东西就是最好到i5-2500k的水准

正。
s********i
发帖数: 17328
5
720p的mkv应该没有问题,hardware decoding。我6G的720p mkv都放过。你可以用
mediainfo贴一下视频信息,看看有没有什么变态的encoding。ATV has limitation of
h.264 decoding - reframe < 16。还要看传输速度,router USB速度很差,另外,如
果用wireless的话,网速也有关系,视频码率高了也可会卡。
a****a
发帖数: 3905
6
来自主题: Hardware版 - 谁在给我讲讲NAS的好处
windows上面twonky,plex,mb3这些server不是一样跑?到底是transcode还是直接
client side decode完全看cilent了。
硬件解码不仅仅是格式支持,码率也要支持。你找个支持hi10p hw decode的出来?
v*****w
发帖数: 1317
7
来自主题: Hardware版 - Video Player for 4K videos
mpc hc 里面试试那些不同的video decoder,再试试hardware decoder 开关

4K
z*********e
发帖数: 10149
8
来自主题: Hardware版 - qnap 653a和三星电视,谁的问题?
doubt that cpu would be enough for decoding.
why not just go samba and let tv/android box decode videos?
a*******t
发帖数: 891
9
yes, need the same software with the same key
use the oldest and most secured encryption method: send a regular mail to her
with a decoding key. send her "encrypted" email which only can be decoded with
your keys : ]
t****h
发帖数: 113
10
来自主题: Java版 - JMF支持DIVX吗?
我试过了,是MPEG-4的decoder,因为一些版权问题,
IBM现在没有发布encoder,但是他们建议可以使用divx的encoder
我自己试过,用IBM的这个decoder,可以播放divx的文件
P*****f
发帖数: 2272
11
来自主题: Java版 - 今天面试。
I remember at least for Jax-RPC, the encoding and decoding interfaces are
proposed, which complex data structure should implement for encoding and
decoding.

当parameter是collection, proxy不会有element 的type info. 所以一个问题是如何保
br />
っ恳桓鰁lement的type to cross platform and language RT. 即使generic也不会解

这个问题.
g****c
发帖数: 299
12
来自主题: Programming版 - 问一道排序题目
special case will overflow
here is mine idea
do runlength coding and special decoding to construct the required sequence.
coding O(n), decoding time complexity average/best caseO(n), I think 2*O(b)~=
O(1) space for runlength need if b is constant.
worst case is tricky, it looks like O(b*(n-b)), could achive O(n) in double for
loops.
g******e
发帖数: 352
13
来自主题: Programming版 - 用python urlopen 抓mitbbs页面的问题
试着用python urlopen 获取mitbbs的页面
Windows下,没有再加encode, decode
遇到一个奇怪的问题,有些帖子的页面抓下来没问题,
但有些帖子的页面获取到的就是乱码,如果用chardet来检测,也检测不出
是什么编码 (返回None). 而那些能正常抓下来的页面,则chardet会正确返回
gb2312
如果加上content.encode('gb2312').encode(type)
就报错'gb2312' codec can't decode bytes in position 1-2: illegal multibyte
sequence
mitbbs所有页面应该都是gb2312编码呀,
有哪位大牛给看看问题出在哪里,谢谢,有包子答谢
g*****y
发帖数: 7271
14
来自主题: Programming版 - 请教一个C++的设计问题
每个decoder register一下,然后loop就好。扩展decoder就很容易了。
N*****m
发帖数: 42603
15
来自主题: Programming版 - 请教一个C++的设计问题
load decoders不是问题
问题是怎么不用重新编译,把header跟decoder对应起来
d**d
发帖数: 389
16
来自主题: Programming版 - 请教一个C++的设计问题
谢谢,这个可行。
我原先准备让factory准备一个static的register函数,然后每个decoder有一个static
的变量。
static boolean decoderXXX::registeStatus=Factory::registerDecoder(std::
function belongToMe, std::function
getInstance)
这样decoder直接就注册了.
S*******e
发帖数: 525
17
来自主题: Programming版 - 再来问问tensorflow的问题
咱这小白看这东西比较吃力。上次多谢wdong及其他几位大师,懂了一点。想听从大师
们的建议,看看tensorflow web site的tutorial,觉得水平不够,还是看看书更好. 现
在又不懂了.这个code:
https://github.com/backstopmedia/tensorflowbook/blob/master/chapters/04_
machine_learning_basics/logistic_regression.py
的输出结果是"不定"的, 我认为是因为38-41行做了sampling:
return tf.train.shuffle_batch(decoded,
batch_size=batch_size,
capacity=batch_size * 50,
min_after_dequeue=batch_size)
我把它改为把所有的数据都读进去:
return tf.train.batch(decoded,
... 阅读全帖

发帖数: 1
18
Decode的时候还是一个个词出,前一个词的hidden layer会feed到后一个词里去。其实
我觉得是脱裤子放屁,老早就有人提过Dilated CNN做encoder, LSTM/GRU做decoder。

overfit

发帖数: 1
19
就是这么做的,但是decoding如果也是CNN的话,不好控制输出长度。比较自然的办法
是CNN encoding + RNN decoding。看狗脸谁先做到更好吧。
c********w
发帖数: 308
20
来自主题: Programming版 - Java 提高performance问题
需要做个简单的Java工具,input(从Unix pipeline 来) 是一千多万行binary msg.
每行被decode 成一个json msg.然后apply 一个filter 最后输出所有matching json
data.
如果decoder 和 filter 已经很优化了。有什么方法或者工具可以提高处理速度?关键
是量太大了。。。

发帖数: 1
21
来自主题: Programming版 - Blackberry的QNX有谁用过么?
這是目前ADAS的Reference Boards,看來QNX要all-in ARM啦,不錯。
Renesas H3 Starter
• OpenGL on GPU
• OpenCV / OpenCL on IMP/IMR cores
• H.264 video encode/decode
Renesas V3M Starter
• OpenCV / OpenCL on IMP and IMR cores
• H.264 encoding
NXP S32V Treerunner
• Native APEX SDK
• OpenCV with APEX acceleration
• OpenCL on GPU
• OpenCL on APEX with limitations
• OpenGL on GPU
Nvidia Drive PX2
• Cuda, CuDNN and several other GPU accelerated technologie... 阅读全帖

发帖数: 1
22
来自主题: Programming版 - Blackberry的QNX有谁用过么?
這是目前ADAS的Reference Boards,看來QNX要all-in ARM啦,不錯。
Renesas H3 Starter
• OpenGL on GPU
• OpenCV / OpenCL on IMP/IMR cores
• H.264 video encode/decode
Renesas V3M Starter
• OpenCV / OpenCL on IMP and IMR cores
• H.264 encoding
NXP S32V Treerunner
• Native APEX SDK
• OpenCV with APEX acceleration
• OpenCL on GPU
• OpenCL on APEX with limitations
• OpenGL on GPU
Nvidia Drive PX2
• Cuda, CuDNN and several other GPU accelerated technologie... 阅读全帖
P***e
发帖数: 804
23
来自主题: Windows版 - [转载] DVD drive怎么放不了DVD?
【 以下文字转载自 Software 讨论区 】
【 原文由 Ponde 所发表 】
说是少个decoder,偶是IBM的机器,哪能当个
decoder啊,随机器也没见啊.谢了现,
还有,哪有好的DVD播放软件啊?
u*******y
发帖数: 8
24
来自主题: XML版 - 请教XML
我正在写一个JAVA对象数据库. 我想提供数据库中的对象和XML之间
的转换, 但不知如何入手.
不会需要自己用REFLECTION来写所有的ENCODE AND DECODE吧.
PRIMITIVE DATA TYPE好办.但OBJECT尤其是REFERENCE可挺麻烦的.
有没有现成的XML-JAVA的PACKAGE呢?
好象XML-RPC 就是用XML来ENCODE AND DECODE JAVA DATA TYPES. 但
偶也不大懂.
ANY IDEA?
THANKS A LOT.
r***y
发帖数: 25
25
It is a quite established field for both neuroscience and engineering
research now. The fundamental was paved by Apostolos P. Georgopoulos (U
Minnesota, http://www.neurosci.umn.edu/faculty/georgopoulos.html) in their papers published in early 1980's about population coding of movement direction in Monkey's motor cortex. Another breakthrough was made by Miguel Nicolelis (Duke U, http://www.nicolelislab.net/) in early 2000's -- manipulate robotic arms/effectors using multi-unit recording in mice a... 阅读全帖
b*****k
发帖数: 80
26
yes,it's a block code
the parity check matrix is very sparse,or said differently, low density.
Gallegar discovered this in 60s and people rediscover this code
in late 90s which shows performance(w/ 10^7 block length)
very close to shannan capacity.
This code and another popular codes--turbo codes use similiar
iterative decoding stragety. A new tool--factor graph
(or normal graph in Forney's vocabulary)
is also formulated to anaylize the behavior of those codes.
LDPC code has lower decoding compl
w*******h
发帖数: 60
27
比如 MIMO decoder or Channel Decoder?
r*****e
发帖数: 620
28
manchester decoder需要一个reference clock作为输入,才能做clock recovery
产生这个reference clock并不是decoder的任务
a******n
发帖数: 293
29
If you only care about the decoder, you may assume all-zero input to the
encoder, which generates all-zero codeword as well. I remember one of my
labmate once memtioned there is a paper claiming that in terms of evaluation
of decoding performance, all-zero input makes no difference with randomized
input sequence. Although I doubt the argument, a lot of people did this for
acdamic research.
c*****t
发帖数: 69
30
来自主题: EE版 - 问个interview问题
是前段的职位。关于memory architecture,请问能否说的具体点?我能想到的方法无
非是用并联多个小的memory,前面加上一个address decoder,decoder和memory之间加
上pipeline。
b*******n
发帖数: 5065
31

it is difficult to find new large prime number.

it is easy and fast to encrypt but so time-consuming to decode.
after you decode it, it is not useful anymore.
u*****o
发帖数: 1224
32
【 以下文字转载自 JobHunting 讨论区 】
发信人: ultrabo (夜空中最亮的星), 信区: JobHunting
标 题: Tower research 面经+ 扭腰一日游
发信站: BBS 未名空间站 (Fri Dec 6 23:05:04 2013, 美东)
面试公司:tower research capital, 职位: quantitative trader, 地点:NYC曼哈顿
这家公司做high frequency trading的,大概就是抄短线的意思。本人对金融业一无所
知,炒过一点股票还高买低卖,但拿到onsite后,还是想着应该锻炼一下自己,面试虽
然痛苦,但肯定是有好处的,何况还可以看看华尔街和金融精英们,都长得啥样。。用
心良苦的把面试安排在感恩节后,想着好好复习,结果跳deal跳的天昏地暗,题也没做
几道。在西雅图候机的时候心情别提多沉重了,onsite丢脸那可是无处遁形啊。怕了一
路,最后还是没躲过。
转机的时候,去纽约的飞机又因为天气问题被cancel了,在芝加哥机场等了3个多小时
。到了纽约已经晚上8点多了。宾馆屋里那叫一个黑呀,根... 阅读全帖
b*****e
发帖数: 223
33
来自主题: Statistics版 - 再求个 sas 的 quick answer
有一些对数据做 decoding 的程序,因为数据比较大,一个个 run 要等。我想把程序
写好,然后用另外的程序去一个个 incl 这些 decoding 程序(有没有更好的方法?)
,晚上我睡觉的时候 run 节省时间。但是我不想因为某个程序出错就 stop 住了,因
为他们之间是无关的。请问怎么写啊。最近比较缺觉,白天想不出来就指望晚上爬上了
问个 quick answer。
m***f
发帖数: 1622
34
看下面的视频,类似的技术正在进行中,
预计不远,就能在实验和数理技术上找到最佳的处理方法,来重新大脑内影像。
http://www.youtube.com/watch?v=daY7uO0eftA
and
Neuron, Volume 60, Issue 5, 915-929, 10 December 2008
Visual Image Reconstruction from Human Brain Activity using a
Combination of Multiscale Local Image Decoders
Yoichi Miyawaki1, 2, 6, Hajime Uchida2, 3, 6, Okito Yamashita2, Masa-aki
Sato2, Yusuke Morito4, 5, Hiroki C. Tanabe4, 5, Norihiro Sadato4, 5 and
Yukiyasu Kamitani2, 3, ,
Summary
Perceptual experience consists of an enormous number of ... 阅读全帖
f*******y
发帖数: 421
35
来自主题: Neuroscience版 - 大规模神经环路建模的必要性
thanks for the link:
Argus II is designed to bypass the damaged photoreceptors altogether. A
miniature video camera housed in the patient’s glasses captures a scene.
The video is sent to a small patient-worn computer (i.e., the video
processing unit – VPU) where it is processed and transformed into
instructions that are sent back to the glasses via a cable. These
instructions are transmitted wirelessly to an antenna in the implant. The
signals are then sent to the electrode array, which emits sm... 阅读全帖
f*******y
发帖数: 421
36
来自主题: Neuroscience版 - 大规模神经环路建模的必要性
thanks for the link:
Argus II is designed to bypass the damaged photoreceptors altogether. A
miniature video camera housed in the patient’s glasses captures a scene.
The video is sent to a small patient-worn computer (i.e., the video
processing unit – VPU) where it is processed and transformed into
instructions that are sent back to the glasses via a cable. These
instructions are transmitted wirelessly to an antenna in the implant. The
signals are then sent to the electrode array, which emits sm... 阅读全帖
o**********e
发帖数: 18403
首页 上页 1 2 3 4 5 6 7 8 9 10 下页 末页 (共10页)