Hello world!

Welcome to WordPress.com. This is your first post. Edit or delete it and start blogging!

Posted in Uncategorized | 1 Comment

Call for Registration: ACM Multimedia 2010 Tutorials

ACM Multimedia 2010 Tutorial Program – Call For Registration 

ACM Multimedia 2010 tutorial program will provide in-depth learning and discussion opportunities on state-of-the-art multimedia research and development.  All tutorials will be held on the first day (Oct. 25, 2010) of the main conference. This year we are offering 14 tutorials that cover current and emerging topics from the four major tracks of the conference. The detailed information of all the tutorials can be found at: http://www.acmmm10.org/program/tutorials/. You can register for any of the tutorials through the main conference registration site at http://www.acmmm10.org/attendees/registration.

Posted in Research | Leave a comment

Interactive Image Search: Search by Color Sketch – Just Released in Microsoft Bing Image Search

On June 3, Microsoft released a new version of its Bing search engine. Among the new functionalities, an interactive image search interface provides a totally new capacity of searching images on the Web. This article will introduce this feature with a couple of interesting examples.

The new feature is based on a research effort from Microsoft Research Asia (MSRA). Last year, the two major contributors for this technique from MSRA, Xian-Sheng Hua and Jingdong Wang, has reported this technology using a number of screenshots of a prototype in MSRA Blog (Chinese version). Now this interesting technique is in Bing image search that everyone can try.

Originally the feature is called “search by color map” in the prototype. But now it is called “Specific color” and “Filter using sketch” in Bing image search. We can call it “image search by color sketch”.

In this article, we will first introduce the steps of using this feature, which has three functionalities, and then present more details and tips of using this feature. And last we will talk a little bit about the technology behind.

Table of Content

How to Use Search by Color Sketch

More Examples

Grass 
Flower 
Flag 
Search people (“Lady Gaga”) 
Search car (“BMW”)

About the Underlying Technologies

 

How to Use Search by Color Sketch

1. Visit http://www.bing.com and click “Images”, or directly visit: http://www.bing.com/images or http://image.bing.com;

2. Check the top-right corner of the web page and change the region of the service to “United States” if currently it is not. Currently this feature is only enabled in US region, but can be used worldwide as long as you changed the region to “United States”;

3. Type a query in the search box and press search button. Now we use “apple” as an example.

4. Click “COLOR” on the left panel (if it is not expanded), then you will see a menu “Specific color”. This is the entrance of the new feature. Currently it is hidden and hope it will be placed somewhere easier to be found (Figure 1).

(a) The whole page

(b) Enlarged: the bottom-right corner of the page

Figure 1: Open the entrance of the new feature.

5. After clicking, you get a panel at bottom-left part of the page (Figure 1), where you can try the first functionality: search by dominant color. That means, you can get “flower” related images but the dominant color of the images would be the one you clicked. For example, if you click the small blue block, you will get results like Figure 2 (mostly blue flowers or flowers’ color close to blue) (Figure 2)

Figure 2: Blue flowers (“zoom small” was clicked so much more images can be shown in one page, which can be set on top-right of the image area)

6. The second and third functionality needs open the drawing panel by checking “Filter using sketch” in the color panel. Then you can choose colors and draw the color sketch of the desired image (Figure 3). Please note this is not for drawing contour of a cat, dog, house or other objects, but only the color “map” of desired images. For example, we may want to find purple flowers with green leaf – that’s what Figure 3 is showing. After drawing the color sketch, press “Apply” to get new search results. We can call this “search by drawing” or “search by color sketch”.

Figure 3: Purple flowers with green leaves.

7. The third functionality could be called “search by modified example”. That means, instead of drawing on the empty panel, we can drag-n-drop an image from the search results to the panel. We can directly click “Apply” to search similar images (Figure 4), or we can draw something to change the example before search. For example, we use the eraser to ease the green leaves and then draw some yellow blocks, which indicates that I’d like to see some yellow followers or parts besides the purple ones (Figure 5).

Figure 4: Search by an example (drag-n-drop an image from the search results).

Figure 5: Search by modified example (purple flowers but with something yellow). The first image in the results will remain the same as the originally selected image no matter how you modified it. This is by design, so people will always know which image he or she has chosen.

There are hyperlinks embedded in the example pictures above (and also those in next section). You can click it to see what it looks like in Bing image search. However, the results may be slightly different (or even much different if you read this article long after June 2010) with the pictures showing here as Bing’s image index is keeping updating. If the links are invalid, you may try it manually using the same term and similar drawing, or any other drawings you like.

Below figure shows the functions of the button in the panel (Figure 6).

Figure 6: Functions of the buttons.

More Examples

We will show you a number of examples, based on which you will learn how to well express you intent through drawing color sketch.

Grass

Type “grass” in Bing image search you will get something like Figure 7.

Figure 7: Original results of “grass” in Bing image search.

If want more sky on the top and less grass at the bottom, you can draw something to express this intent and get something like Figure 8.

Figure 8: “Grass” images with more sky on top and less grass at bottom.

Or you want less sky on top and more grass at bottom (Figure 9):

Figure 9: “Grass” images with little sky and lots of grass.

Sometimes you may want to find “orange grass” (Figure 10), or you find one of the orange grass looks more like what you want (drag it to the panel as an example) (figure 11).

Figure 10: Orange grass.

Figure 11: Search by example “grass”

Sometimes, you may want to get some grass images with clean background, so it will be easier to use it. This can be achieved by set the background to white or black or any other possible background color (Figure 12).

Figure 12: Grass images with clean/white background

Another example of searching by example: grass with orange sky (Figure 13), and modified example (similar grass with blue sky) (Figure 14).

Figure 13: Search by example (grass with orange sky area).

Figure 14: Search by modified example (similar grass but with blue sky). Again, the first image in the results will remain the same as the originally selected image no matter how you modified it. This is by design, so people will always know which image he or she has chosen.

Flower

Original results of searching “flower” in Bing image search (Figure 15). Blue flowers or some flower related examples have been shown in Figure 2, 3, 4 and 5. Here are more examples: purple flower (Figure 16), yellow flower (Figure 17), yellow flower with dark background (Figure 18), search by example (an yellow flower with dark background) (Figure 19), and search by modified example (yellow flower with green background)(Figure 20).

Figure 15: Original results of “flower” in Bing image search.

Figure 16: Purple flowers (by selecting the dominant color).

Figure 17: Yellow flowers (indicated by yellow blocks in the center area).

Figure 18: Search by drawing (yellow flowers with dark background, indicated by yellow blocks at center area and black blocks at the border).

Figure 19: Search by example (yellow flower with dark background).

Figure 20: Search by modified example (change the background to green, indicating leaves or grass).

Flag

When you saw a flag of a country when watching world cup, but you don’t know what the country of the flag, you may use Bing image search to find it out. Below shows several results (Figure 21 to Figure 27) – after drawing the flag roughly, you get the real flag images. Mouse-over or click the desired result, you will know the country of the flag.

Figure 21: China flag.

Figure 22: India flag (and other similar flags, such as Miami city flag and Niger flag)

Figure 23: Niger flag (and other similar ones).

Figure 24: Ghana flag.

Figure 25: Mauritius flag (search by example).

Figure 26: Gambia flag (search by modified example).

Figure 27: Danmark flag and similar flags (search by drawing).

My son (2.5 years old) is able to recognize more than 80 flags of different countries, but I can only recognize 10+. Now with Bing’s “search by color sketch”, I can compete with him now, thought the speed is much slower than him 🙂

Search people (“Lady Gaga”)

“Lady Gaga” is very popular recently. Figure 28 shows the initial results in Bing. Figure 29, 30, 31, and 32 shows the images of Lady Gata walking on red carpet and wearing in different colors.

Figure 28: Original results of “Lady Gaga” from Bing image search.

Figure 29: Lady Gaga walking on red carpet (indicated by a red line at the bottom).

Figure 30: Lady Gaga walking on red carpet and wearing in black (indicated by a red line at the bottom and some black blocks near the center.

Figure 31: The same as above but wearing in white (sometimes you need to adjust the size and/or position of the color blocks to get better results).

Figure 32: Walking on red carpet while background is white.

Figure 33: Search by an example (wearing in black and blue).

Figure 34: Dark background.

Figure 35: Dark background with larger face area.

Figure 36: Search by another example with dark background.

Search car (“BMW”)

Figure 37 to Figure 43 show a couple examples for search color of different colors and different background.


Figure 37: Original results of “BMW” of Bing.

Figure 38: Blue BMW.

Figure 39: Red BMW.

Figure 40: Red BMW on lawn.

Figure 41: Red BWM on lawn with blue sky.

Figure 42: Search by modified example. Originally the car is black (i.e., the first image in the result list). Now it was changed to red – we got red cars in similar background.

Figure 43: The same as the above example, but we change the color of the car to orange.

About the Underlying Technologies

This is not re-ranking, not conventional “query by sketch”, not just user-interface. It should be the first web-scale “image search by drawing” system. When drawing is provided, a new query based on the text query and drawing will be issued to search new images from the entire Bing index.

……

(To be continued – some high-level introduction of the technology and the differences compared with existing work, as well as some tips of using the feature.)

By Xian-Sheng Hua, Beijing, June 2010

Posted in Research | 8 Comments

Microsoft Academic Search (paper search)

http://academic.research.microsoft.com/

 Try it. Still in beta.

 (I was asked to give a try by some people working on it :))

Pros:

Good UI. Support author search/rank list, conference search/rank list, journal search/rank list, and some other cool features (try it J)

For example, top ranked authors in multimedia area:

http://academic.research.microsoft.com/CSDirectory/Author_category_13.htm  (hope it was done automatically)

And it also support author profile page, for example: http://academic.research.microsoft.com/Author/951346.aspx 

However, the big problem is it cannot differentiate different people with the same name either. But seems there is no a search engine can solve this problem currently.

 

Cons:

Dataset is not complete enough. Missing papers, citations, and indexes (such as g-index and h-index) are much blow the correct ones.

For the nice feature mentioned above (author profile), the downside is that it also provides a photo of the author, however, the listed papers may come from different people with the same name. The person whose papers are listed but the photo is not his/her may get angry about it 🙂

 

I believe they are improving, very quickly.I heard they are updating data every week and updating features every two months.

Posted in Research | 1 Comment

[转]青菜的味道

早上上班路上听了一个广播故事,颇有同感。

【播文】
   有一个农民兄弟这样感慨:我们好不容易吃上肉,你们城里人又改吃青菜了。国外也有一种流行说法:穷人吃肉,富人吃青菜。居城多年,聚餐时有一个细节让我记忆深刻──在合上点菜单的时候,宾主间总会有人补充一句﹕“来个青菜吧,最好是带叶子的那种。”

  青菜,神奇的青菜,有时声势浩大地在贫富和城乡之间划一道分界线,有时仅仅是一桌菜品中微不足道的小配角。不管贫富,也不论城乡,过日子终归是离不开青菜的,念念青菜生活味。

  有一次,朋友问我:“你知道青菜的味道吗﹖”第一反应竟是茫然,缓过神来,怀疑他在给我开类似脑筋急转弯的玩笑。我敛起笑容,煞是认真地戏谑道﹕“青菜的味道?不就是青菜味道嘛!”

  朋友却一脸认真地说﹕“我刚刚从成都回来。川菜久负盛名,那些日子我算是真正领略到了。印象最深的不是大鱼大肉,麻辣水煮,而是一家并不起眼的小馆子里的青菜。那家餐馆名字也很有意思,叫青菜人家,土得掉渣,像个裹着头巾羞涩的村姑。进得店来,迎面一张屏风写了一句──你知道青菜的味道吗﹖”

  我等不及,反问道﹕“那你知道青菜的味道吗﹖”朋友说:“从那家店里出来,就知道了。青菜人家有一道水煮青菜,不加任何着料,在滚水里过一遍,就端上桌。吃起来,才是真正的青菜味道。那味道其实很平淡,有些涩口,带点草香,下肚后有回甘,就像割草机修剪过草坪之后,空气里流淌的那味儿了。”

  我说:“不稀奇,国外有人吃青菜都不用在开水过一遍,生吃。那才是叫原味呢。”

  朋友说:“你说得不对,他们吃生的青菜,都要蘸五花八门的酱呢。”

  蓦然间,我想起小时候吃的青菜来。那时,家里做饭用大木甑蒸,饭熟了,母亲会将洗净的青菜──印象中空心菜占多数,放进蒸饭的滚水一焯,洒上一小勺盐,就直接下饭。儿时吃的青菜,没有酱醋等诸般纠缠,没有经烈火锅油炙烤,不走偏,不失真,吃进嘴里,是青菜的原味。只有吃过这样的青菜,才有资格回答“青菜是什么味道”这一并不复杂的问题。可是,红尘中的你我,有几人吃过这样清淡寡味的青菜呢?

  多年来,我们盘中和嘴里的青菜,都被旺火开发过,被油盐酱醋浸润过,其味都被各种杂味抢了风头,遮盖了去,真味反而模糊起来,记不清晰了。

  青菜味道,恰如人生万般滋味,缠绕在味蕾上的是理不清道不明的枝枝蔓蔓,原味和真味,往何处寻觅呢?

  人生之初,我们如吃寡淡的原味青菜。当然,谁也不愿意长久地这样淡下去,所以,不断地树立自己远大的人生目标,苦苦奋斗,孜孜以求。

  等梦想实现了,就如我们吃那被各种调料包围的青菜,原味道于种种干扰中,模糊了,失偏了。正所谓人生百味。老之将至,万事放下,心头了无挂碍,经过百般滋味的历练,终究回归至原味,就像我那朋友从“青菜人家”那里所尝到的原味青菜,就像我儿时吃的那滚水里焯一遍的青菜。

  青菜的味道,从另一角度看,是微版的百味人生。

  人们总在努力追求着,苦苦追寻着,却在不断追求中,丢失了自己最初的目标。等到我们想起那时本真的愿望,回味当初的向往,才发现事过境迁人渐老。

  走在人生道路上,走着走着,不知不觉把自己走丢了,迷失在诱惑、欲望、惊喜、烦恼、痛苦和悔恨等杂陈的人生滋味中。老子曰:“大音希声,大象无形。”我想补充一句:大味至淡。“你知道青菜的味道吗?”一语惊心,不由地在内心自问──你知道成长的味道吗?你知道读书的味道吗?你知道爱的味道吗?你知道……

  味道的迷失,似乎存在于人生的每一时,每一处,处在迷失的当头,各种本真的原味对我们来说,有时竟会是那么奢侈。

  (摘自香港《大公报》 作者:志宏)

【简评】

与《菜根谭》上的这句可互相印证:“甘辛膿肥非真味,真味只是淡,神奇卓異非至人,至人只是常。”不要忘了自己最初的目标,不要迷失在为达到目标而选择路径上 – 那还可能不是最好的路径。

《老子》第四十一章原文:“上士闻道,勤而行之;中士闻道,若存若亡;下士闻道,大笑之。不笑不足以为道。故建言有之: 明道若昧;进道若退;夷道若纇;上德若谷;大白若辱;广德若不足;建德若偷;质真若渝;大方无隅;大器晚成;大音希声;大象无形;道隐无名。夫唯道,善贷且成。”

Posted in Uncategorized | 3 Comments

“查颜观色”

最近为MSRA博客写的一篇短文,介绍基于颜色的图像搜索:
 
 
欢迎大家去踩踩
Posted in Research | 4 Comments

做学问当效季老

早起上班路上听广播,收音机里讲了刚刚辞世的季羡林老先生关于“国学大师”和“国宝”的几段文字,感触颇深 — 做学问当效季老:
 
辞“国学大师”:
 
    现在在某些比较正式的文件中,在我头顶上也出现“国学大师”这一灿烂辉煌的光环。这并非无中生有,其中有一段历史渊源。

  约摸十几二十年前,中国的改革开放大见成效,经济飞速发展。文化建设方面也相应地活跃起来。有一次在还没有改建的北京大学大讲堂里开了一个什么会,专门向同学们谈国学。当时主席台上共坐着五位教授,每个人都讲上一通。我是被排在第一位的,说了些什么话,现在已忘得干干净净。一位资深记者是北大校友,在报上写了一篇长文《国学热悄悄在燕园兴起》。从此以后,其中四位教授,包括我在内,就被称为“国学大师”。他们三位的国学基础都比我强得多。他们对这一顶桂冠的想法如何,我不清楚。我自己被戴上了这一顶桂冠,却是浑身起鸡皮疙瘩。

  说到国学基础,我从小学起就读经书、古文、诗词。对一些重要的经典著作有所涉猎。但是我对哪一部古典,哪一个作家都没有下过死功夫,因为我从来没想成为一个国学家。后来专治其他的学术,浸淫其中,乐不可支。除了尚能背诵几百首诗词和几十篇古文外;除了尚能在最大的宏观上谈一些与国学有关的自谓是大而有当的问题比如天人合一外,自己的国学知识并没有增加。环顾左右,朋友中国学基础胜于自己者,大有人在。在这样的情况下,我竟独占“国学大师”的尊号,岂不折煞老身(借用京剧女角词)!我连“国学小师”都不够,遑论“大师”!

  为此,我在这里昭告天下:请从我头顶上把“国学大师”的桂冠摘下来。

 
     (见季羡林《病榻杂记》)。
 
辞“学界泰斗”
 
  这要分两层来讲:一个是教育界,一个是人文社会科学界。

  先要弄清楚什么叫“泰斗”。泰者,泰山也;斗者,北斗也。两者都被认为是至高无上的东西。

  光谈教育界。我一生做教书匠,爬格子。在国外教书10年,在国内57年。人们常说:“没有功劳,也有苦劳。”特别是在过去几十年中,天天运动,花样翻新,总的目的就是让你不得安闲,神经时时刻刻都处在万分紧张的情况中。在这样的情况下,我一直担任行政工作,想要做出什么成绩,岂不戛戛乎难矣哉!我这个“泰斗”从哪里讲起呢?

  在人文社会科学的研究中,说我做出了极大的成绩,那不是事实。说我一点成绩都没有,那也不符合实际情况。这样的人,滔滔者天下皆是也。但是,现在却偏偏把我“打”成泰斗。我这个泰斗又从哪里讲起呢?

  为此,我在这里昭告天下:请从我头顶上把“学界(术)泰斗”的桂冠摘下来。

 
辞“国宝”
 
  在中国,一提到“国宝”,人们一定会立刻想到人见人爱憨态可掬的大熊猫。这种动物数量极少,而且只有中国有,称之为“国宝”,它是当之无愧的。可是,大约在八九十来年前,在一次会议上,北京市的一位领导突然称我为“国宝”,我极为惊愕。到了今天,我所到之处,“国宝”之声洋洋乎盈耳矣。我实在是大惑不解。当然,“国宝”这一顶桂冠并没有为我一人所垄断。其他几位书画名家也有此称号。

  我浮想联翩,想探寻一下起名的来源。是不是因为中国只有一个季羡林,所以他就成为“宝”。但是,中国的赵一钱二孙三李四等等,等等,也都只有一个,难道中国能有13亿“国宝”吗?

  这种事情,痴想无益,也完全没有必要。我来一个急刹车。

  为此,我在这里昭告天下:请从我头顶上把“国宝”的桂冠摘下来。

 
媒体赞季老:虚怀若谷 对虚荣社会无声抗议 (原文
 
  香港大公报文章指出,每每说到季羡林老先生,国人无不敬重有加,这不仅因为他渊博的知识,他不但精通几十种外文,对一些已经面临绝迹的文字也颇有研究,尤其对中世纪的印欧语言的研究达到了相当的水平,同时,他平实的为人更是令人称道。近几天中央电视台播放的“2006感动中国”的节目中,有一组对季老先生的采访镜头令人难忘。有一个掏粪工是一个业余画家,他出了自己的画集之后想请名家作序,找了几个颇有点名气的人作序时,人家都以种种理由予以回绝。而我们的季老先生却没有嫌弃这位掏粪工,慷慨应允作序。一滴水可以见太阳,从这件小事中人们可以领略季老先生的人生风范。

  文章指出,国学大师、学界泰斗、国宝,这些含金量极高令人羡慕的字眼,是多少人梦寐以求的,而在季老那里却十分地看轻。可以说,季老先生以自己的实际行动给那些追捧虚荣的人上了很好的一课。三项诱人的桂冠就这样被我们的季老先生“轻易”地自己给自己摘去了,这在很多人看来是多么地不可思议!要知道,那是很多人终身奋斗的目标啊!这也充分反映了季老先生“视功名如过眼烟云”的坦荡胸怀。

  季羡林的“请辞”大有深意,这样的“请辞”是对贪慕虚荣社会的一种无声的抗议!给我们的每个社会中人上了很好的一课。论才学也好,论人品也好,季老都当之无愧,可他老人家却并没有如此地“看重”自己,而是把自己当成一个普通的百姓,这样的胸怀正是我们这个社会所缺少的一种精神资源。面对物欲横流的社会,很多人开始丧失自己的道德操守,在学术界更是上演了一幕幕肮脏的“闹剧”,为了获得高级职称,不惜造假,为了显示学术成果,不惜剽窃别人的论文,所有这些都严重地败坏着学界风气。不但如此,在其它领域,贪慕虚荣,崇尚奢华的风气日甚一日,如果我们的社会被这样一种虚假的气氛所笼罩,国家未来的前途令人堪忧。

  社会文明和进步的重要标志是崇尚实干,而不是欺上瞒下,欺世盗名。这方面,季老先生给国人树立了一个很好的榜样。愿世人能从季老先生的“请辞”中得到教益。

  香港文汇报就此还采访了季羡林先生。

  记 者:现在社会上和学术界对您有些微词。
  季羡林:有微词好。
  记 者:您为什么要摘掉自己头上的光环?像是“国宝”、“泰斗”、“大师”这些称号?
  季羡林:我觉得自己不够格,所以想摘掉。
  记 者:您觉得谁够格?
  季羡林:有很多。(迟疑了一下)鲁迅是一个。在那个时代,鲁迅的骨头是最硬的。
  记 者:您对这个民族做出的贡献,是有目共睹的。
  季羡林:我在国外多年,回来后做了一点事,给我的荣誉太高太多,我担受不起。我的人生轨迹是直线向上的,很简单。

Posted in Research | Leave a comment

[转]养成让自已进步的26个习惯

养成让自已进步的26个习惯

一. 永远不说三个字“不可能”。
二. 凡事第一反应是找方法,而非是找借口
三. 遇到挫折时大声对自己说:“太棒了,我终于有机会成长了。”
四. 不说消极的话,不落入消极的情绪当中,一旦出现问题应正面处理。
五. 凡事先定订目标,并尽量制作梦想版。
六. 凡事预先做计划,尽量将目标视觉化。
七. 是工作和学习的时间,就要全部的用在工作和学习上,不要盗用工作和学习的时间。
八. 养成记录的习惯,不要太依赖自己的脑袋记忆。
九. 随时记录灵感。
十. 把重要的观点,目标,方法写下来,并贴出来,随时提醒自己。
十一.走路比平时快30%,走路是脚尖用力向前推进,肢体语言健康有力,不懒散,不颓废。
十二.每天出门照镜子,给自己一个自信的微笑。
十三.每天自我反省一下,自检一下。
十四.开会时坐前排。
十五.时时刻刻微笑待人处事。
十六.用心倾听,不打断别人的话,作一个倾听高手。
十七.说话声声有力,能感染自己,能产生磁场。
十八.同理心,说话之前先考虑对方的感受。
十九. 每天有意识的真诚的赞美别人三次以上。
二十.  及时写感谢卡。
二十一.用关心和自责的口吻说话,责人之前先责己。
二十二.每天进步一点点,日有一新,月有一进,每天多做一件事。
二十三.提前上班,推迟下班。
二十四.节俭并定期存钱。
二十五.时常运用头脑风暴。
二十六.遵守诚信,说到做到。

Posted in Uncategorized | Leave a comment

[转贴]职场赢家不可不知的十个定律

想由“职场菜鸟”练就成“职场高手”并不是一蹴而就的,靠着蛮干提高工作效率显然也很不靠谱。因此,我收集并总结了十条职场定律给朋友们,希望对大家的工作能够有所帮助。

  1. 彼得原理

  每个组织都是由各种不同的职位、等级或阶层的排列所组成,每个人都隶属于其中的某个等级。彼得原理是美国学者劳伦斯·彼得在对组织中人员晋升的相关现象研究后,得出一个结论:在各种组织中,雇员总是趋向于晋升到其不称职的地位。彼得原理有时也被称为“向上爬”的原理。

  这种现象在现实生活中无处不在:一名称职的教授被提升为大学校长后,却无法胜任;一个优秀的运动员被提升为主管体育的官员,而无所作为。

  对一个组织而言,一旦相当部分人员被推到其不称职的级别,就会造成组织的人浮于事,效率低下,导致平庸者出人头地,发展停滞。

  因此,这就要求改变单纯的“根据贡献决定晋升”的企业员工晋升机制,不能因某人在某个岗位上干得很出色,就推断此人一定能够胜任更高一级的职务。将一名职工晋升到一个无法很好发挥才能的岗位,不仅不是对本人的奖励,反而使其无法很好发挥才能,也给企业带来损失。

  2. 酒与污水定律

  酒与污水定律是指把一匙酒倒进一桶污水,得到的是一桶污水;如果把一匙污水倒进一桶酒,得到的还是一桶污水。在任何组织里,几乎都存在几个难弄的人物,他们存在的目的似乎就是为了把事情搞糟。最糟糕的是,他们像果箱里的烂苹果,如果不及时处理,它会迅速传染,把果箱里其他苹果也弄烂。

  “烂苹果”的可怕之处,在于它那惊人的破坏力。一个正直能干的人进入一个混乱的部门可能会被吞没,而一个无德无才者能很快将一个高效的部门变成一盘散沙。组织系统往往是脆弱的,是建立在相互理解、妥协和容忍的基础上的,很容易被侵害、被毒化。

  破坏者能力非凡的另一个重要原因在于,破坏总比建设容易。一个能工巧匠花费时日精心制作的陶瓷器,一头驴子一秒钟就能毁坏掉。如果一个组织里有这样的一头驴子,即使拥有再多的能工巧匠,也不会有多少像样的工作成果。如果你的组织里有这样的一头驴子,你应该马上把它清除掉,如果你无力这样做,就应该把它拴起来。

  3. 木桶定律

  水桶定律是讲一只水桶能装多少水,这完全取决于它最短的那块木板。这就是说任何一个组织,可能面临的一个共同问题,即构成组织的各个部分往往是优劣不齐的,而劣势部分往往决定整个组织的水平。

  “水桶定律”与“酒与污水定律”不同,后者讨论的是组织中的破坏力量,“最短的木板”却是组织中有用的一个部分,只不过比其他部分差一些,你不能把它们当成烂苹果扔掉。强弱只是相对而言的,无法消除,问题在于你容忍这种弱点到什么程度,如果严重到成为阻碍工作的瓶颈,你就不得不有所动作。

  4. 马太效应

  《新约·马太福音》中有这样一个故事:一个国王远行前,交给3个仆人每人一锭银子,吩咐道:“你们去做生意,等我回来时,再来见我。”国王回来时,第一个仆人说:“主人,你交给我的一锭银子,我已赚了10锭。”于是,国王奖励他10座城邑。第二个仆人报告:“主人,你给我的一锭银子,我已赚了5锭。”于是,国王奖励他5座城邑。第三仆人报告说:“主人,你给我的1锭银子,我一直包在手帕里,怕丢失,一直没有拿出来。”于是,国王命令将第三个仆人的1锭银子赏给第一个仆人,说:“凡是少的,就连他所有的,也要夺过来。凡是多的,还要给他,叫他多多益善。”这就是“马太效应”,反应当今社会中存在的一个普遍现象,即赢家通吃。

  对企业经营发展而言,马太效应告诉我们:要想在某一个领域保持优势,就必须在此领域迅速做大。当你成为某个领域的领头羊时,即便投资回报率相同,你也能更轻易地获得比弱小的同行更大的收益。而若没有实力迅速在某个领域做大,就要不停地寻找新的发展领域,才能保证获得较好的回报。

  5. 零和游戏原理

  零和游戏是指一项游戏中,游戏者有输有赢,一方所赢正是另一方所输,游戏的总成绩永远为零,零和游戏原理之所以广受关注,主要是因为人们在社会的方方面面都能发现与“零和游戏”类似的局面,胜利者的光荣后面往往隐藏着失败者的辛酸和苦涩。

  20世纪,人类经历两次世界大战、经济高速增长,科技进步、全球一体化以及日益严重的环境污染,“零和游戏”观念正逐渐被“双赢”观念所取代。人们开始认识到“利已”不一定要建立在“损人”的基础上。通过有效合作皆大欢喜的结局是可能出现的。

  但从“零和游戏”走向“双赢”,要求各方面要有真诚合作的精神和勇气,在合作中不要小聪明,不要总想占别人的小便宜,要遵守游戏规则,否则“双赢”的局面就不可能出现,最终吃亏的还是合作者自己。

  6. 华盛顿合作规律

  华盛顿合作规律说的是一个人敷衍了事,两个人互相推诿,三个人则永无成事之日。多少有点类似于我们“三个和尚”的故事。

  人与人的合作,不是人力的简单相加,而是要复杂和微妙得多。在这种合作中,假定每个人的能力都为1,那么,10个人的合作结果有时比10大得多,有时,甚至比1还要小。因为人不是静止物,而更像方向各异的能量,相互推动时,自然事半功倍,相互抵触时,则一事无成。

  我们传统的管理理论中,对合作研究得并不多,最直观的反映就是,目前的大多数管理制度和行为都是致力于减少人力的无谓消耗,而非利用组织提高人的效能。换言之,不妨说管理的主要目的不是让每个人做得更好,而是避免内耗过多。

  7. 手表定理

  手表定理是指一个人有一只表时,可以知道现在是几点钟,当他同时拥有两只表时,却无法确定。两只手表并不能告诉一个人更准确的时间,反而会让看表的人失去对准确时间的信心。

  手表定理在企业经营管理方面,给我们一种非常直观的启发,就是对同一个人或同一个组织的管理,不能同时采用两种不同的方法,不能同时设置两个不同的目标,甚至每一个人不能由两个人同时指挥,否则将使这个企业或这个人无所适从。

  手表定理所指的另一层含义在于,每个人都不能同时选择两种不同的价值观,否则,你的行为将陷于混乱。

  8. 不值得定律

  不值得定律最直观的表述是:不值得做的的事情,就不值得做好。这个定律再简单不过了,重要性却时时被人们忽视遗忘。不值得定律反映人们的一种心理,一个人如果从事的是一份自认为不值得做的事情,往往会保持冷嘲热讽,敷衍了事的态度,不仅成功率低,而且即使成功,也不觉得有多大的成就感。

  因此,对个人来说,应在多种可供选择的奋斗目标及价值观中挑选一种,然后为之奋斗。“选择你所爱的,爱你所选择的,才可能激发我们的斗志,也可以心安理得。而对一个企业或组织来说,则要很好地分析员工的性格特性,合理分配工作,如让成就欲较强的职工单独或牵头完成具有一定风险和难度的工作,并在其完成时,给予及时的肯定和赞扬;让依附欲较强的职工,更多地参加到某个团体中共同工作;让权力欲较强的职工,担任一个与之能力相适应的主管。同时要加强员工对企业目标的认同感,让员工感觉到自己所做的工作是值得的,这样才能激发职工的热情。

  9. 蘑菇管理

  蘑菇管理是许多组织对待初出茅庐者的一种管理方法,初学者被置于阴暗的角落(不受重视的部门,或打杂跑腿的工作),浇上一头大粪(无端的批评、指责、代人受过),任其自生自灭(得不到必要的指导和提携)。相信很多人都有过这样一段“蘑菇”的经历,这不一定是什么坏事,尤其是当一切刚刚开始的时候,当几天“蘑菇”,能够消除我们很多不切实际的幻想,让我们更加接近现实,看问题也更加实际。

  一个组织,一般对新进的人员都是一视同仁,从起薪到工作都不会有大的差别。无论你是多么优秀的人才,在刚开始的时候,都只能从最简单的事情做起,“蘑菇”的经历,对于成长中的年轻人来说,就象蚕茧,是羽化前必须经历的一步。所以,如何高效率地走过生命的这一段,从中尽可能汲取经验,成熟起来,并树立良好的值得信赖的个人形象,是每个刚入社会的年轻人必须面对的课题。

  10. 奥卡姆剃刀定律

  12世纪,英国奥卡姆的威廉主张唯名论,只承认确实存在的东西,认为那些空洞无物的普遍性概念都是无用的累赘,应当被无情地“剃除”。他主张“如无必要,勿增实体”。这就是常说的“奥卡姆剃刀”。这把剃刀曾使很多人感到威胁,被认为是异端邪说,威廉本人也因此受到迫害。然而,并未损害这把刀的锋利,相反,经过数百年的岁月,奥卡姆剃刀已被历史磨得越来越快,并早已超载原来狭窄的领域,而具有广泛、丰富、深刻的意义。
 
Posted in Uncategorized | 1 Comment

同学们,踊跃投稿啊 – ICDM 2009 Workshop: Internet Multimedia Mining

同学们,踊跃投稿啊, 顶啊,帮忙宣传一下啦 

 

 

1st International Workshop on Internet Multimedia Mining
(with dataset support)

In conjunction with IEEE International Conference on Data Mining 2009

December 6 2009, Miami, Florida, USA

 

Aims and scope

 

With the explosion of video and image data available on the Internet, online multimedia applications become more and more important. Moreover, mining semantics and other useful information from large-scale Internet multimedia data to facilitate online and local multimedia content analysis, search and other related applications has also gained more and more attention from both academia and industry. On the one hand, the rapid increase of online multimedia data brings new challenges to multimedia content analysis, multimedia retrieval and related multimedia applications, especially in scalability. Both computation cost and performance of many existing techniques are far from satisfactory. On the other hand, Internet also provides us with new opportunities to attack these challenges as well as conventional problems encountered in multimedia mining, content analysis, image/video understanding and computer vision. That is, the massive associated metadata, context and social information available on the Internet, as well as the massive grassroots Internet users, are invaluable resources that can be leveraged to solve the aforementioned difficulties.

 

Recently more and more researchers are realizing both the challenges and the opportunities for multimedia research brought by the Internet. This workshop aims at bringing together high-quality and novel research works on "Internet Multimedia Mining".

 

One of the major obstacles of "Internet Multimedia Mining" research is the difficulty in  forming a "good" dataset for algorithm developing, system prototyping and performance evaluation. Together with this workshop, we release a benchmark dataset, which is based on real Internet multimedia data and real Internet multimedia search engines. Submissions to this workshop are encouraged to use this dataset, but papers/demos working on other Internet-based datasets are also welcome.
 

MSRA Multimedia Dataset

 

MSRA-MM Version 1 dataset is ready for shipping. Detailed information about the dataset can be find here. Please contact the dataset chair (Meng Wang) to request the data.

 

MSRA-MM Version 2 dataset (10 times larger with more metadata) will be ready around June 15. Please submit your request to the dataset chair.
 

Please be noted that using these datasets is NOT mandatory. Papers/demos working on other Internet-based datasets are also welcome.

Topics of Interest

 

Topics of interest for this workshop include, but are not limited to:

  • Internet video/image/audio annotation, classification, tagging, search ranking and reranking by combining textual description and video/image content.

  • General video/image/audio annotation, classification, tagging, search ranking and reranking by exploiting Internet data and/or users. Approaches which can handle large-scale data/users are more preferred.

  • Video/image/audio processing and analyses using Internet data as a knowledge base.

  • Social media processing, such as online media authoring and sharing, tag recommendation, tag filtering, tag ranking, and search ranking based on image/video/audio social context.

  • Knowledge mining from Internet multimedia data, such as mining semantic distance of keywords or images, mining video/image/audio copy relationships (e.g., given a video/image/audio, to find all videos/images on the Internet that have the same content with the video/image, either entirely or partially), mining trends of multimedia consuming/sharing, mining knowledge (for example, "photo encyclopedia"", from massive amount of multimedia content on the Internet, etc.

  • Web-scale content-based multimedia retrieval (for example, approaches based on large-scale high-dimensional feature indexing).

  • Other online multimedia mining applications, such as multimedia advertising, multimedia recommendation, as well as location/GPS/geography-enabled multimedia, multimedia sensor network over the Internet, etc.

 

Paper Submission

We accept two forms of submissions: regular full papers and demonstrations. Regular submissions for this workshop are required to use the same format as regular ICDM long papers (a maximum of 10 pages in the IEEE 2-column format). And demonstration submission requires a 1- or 2-page demo description. We especially encourage long-paper authors to submit a demo also. All submissions will be peer-reviewed by at least 3 members of the program committee. Extended version of selected papers will be invited to submit to a special issue of a top journal in data mining or multimedia area.

 

Paper submission entrance

 

 CMT Conference System

 

Awards

 

The workshop will present two awards: a best paper award and a best demonstration award, judged by a separate awards committee.

 

Important Dates

  • August 8 Submission of full paper

  • August 29 Submission of demo paper

  • September 8 Notification of Acceptance

  • September 28 Camera-Ready Paper Due

  • December 6 Workshop

 Workshop Co-Chairs
   
     Xian-Sheng Hua, Microsoft Research Asia, China
        Cees G.M. Snoek, University of Amsterdam, The Netherlands
        Zhi-Hua Zhou, Nanjing University, China

Dataset Chair
 
    Meng Wang, Microsoft Research Asia, China

 

Program Committee
        Shih-Fu Chang, Columbia University, USA

        Lingyu Duan, Peking University, China

        Alan Hanjalic, Delft University of Technology, the Netherlands

        Winston Hsu, National Taiwan University, Taiwan

        Yiannis Kompatsiaris, Informatics and Telematics Institute, Greece

        Shipeng Li, Microsoft Research Asia, China

        Zhu Li, Hong Kong Polytechnic University, Hong Kong

        Jiebo Luo, Kodak Research, China

        Stéphane Marchand-Maillet, University of Geneva, Switzerland

        Chong-Wah Ngo, City University of Hong Kong, Hong Kong

        Shin’ichi Satoh, National Institute of Informatics, Japan

        Nicu Sebe, University of Amsterdam, the Netherlands

        Dacheng Tao, Nanyang Technological University, Singapore

        Marcel Worring, University of Amsterdam, the Netherlands

        Changsheng Xu, Chinese Academy of Sciences, China

        Zhongfei Zhang, Binghamton University, USA

       

        (To be completed)

Posted in Research | 3 Comments