To contact the students of my exam-exempt graduate student-Shanshi researcher-Institute of Computational Sciences

To contact the students who are enrolled in my exam

Shanshi

August 23, 2012 (revised)

Good Fellow students: First of all, it's a pleasure to be acquainted with you. Each year with the exemption of the recommended graduate students are many, you contacted me, for whatever reason, we have a chance to become a teacher and student fate. But unfortunately, I usually only recruit 2 graduate students per year (or straight Berson), and contact my classmates every year more than 40 people (of course, not these students will eventually report me), So, "choose" to become me and you have to "common choice", the reason is said "common choice", because the process itself is not " I choose you "so simple, and more also includes" your choice of me. " In order to better let you know me and make your "real choice", I wrote down this document. The document is a bit long, but I suggest you read this article carefully, I believe it will be of benefit to you. Links in this document: My team, my research interests, my scientific philosophy. You may not be suitable for my team to join my team what can you get about the interview Research team in the area If you end up being my graduate student, it doesn't just mean that we become a teacher-student relationship, it also means that you're going to be part of my research team, so it's recommended that you look at my research team first. From the point of view of administrative relations, I am a researcher and doctoral tutor of the Institute of Computing Technology of CAS, and are affiliated with the visual Information processing and Learning Research Group (VIPL) of the Chinese Academy of Sciences Key Laboratory for Intelligent Information Processing (IIP Lab). IIP Lab, a key laboratory of the CAS, is one of the two laboratory-level key labs in the institute, while my close collaborators for many years Chen Xilin fellows are currently the Director of the key laboratory, and I am currently the deputy Director of the laboratory. The VIPL Study Group currently has three associate researchers (Changhong, Miaojun, Chang) and four research assistants (Chai Shujuan, Wang Ruiping, Wang Yu, Lambert Qin) and several postdoctoral fellows, in addition to Mr. Chen and me. In addition, he is still a visiting researcher at the VIPL Research group. The main research direction of VIPL is digital media information processing and understanding, especially intelligent human-computer interface technology, AV codec, visual computing and recognition, multimedia understanding and retrieval. In addition, our research team is logically affiliated to my doctoral mentor, Professor Gao Gao, a joint laboratory of advanced human-machine communication technology (JDL), JDL is specialized in multimedia technology and intelligent human-computer Interaction technology research institutions, Research and development goals are mainly focused on the core technology of broadband network multimedia system, and the development of key application systems. The main researchers are from the Institute of Digital Media Research center of CAS, graduate School of Chinese Academy of Sciences, Beijing University of Digital Media, School of Computer, Harbin Institute of Technology, School of Computer science, and other universities and industry from home and abroad. The research fields of the joint laboratory include the digital audio and video codec technology, the content-based mass multimedia information retrieval, the human biometric detection and recognition, the intelligent human-computer interaction technology, the application of the law and other five directions. The joint laboratory currently undertakes a number of research topics such as the national "973" program, the National Natural Science Fund, the National "863" program, the National Eleven-Five Science and Technology Support program, and many international cooperation projects. After several years of efforts, the joint laboratory has achieved a number of innovative research results, has won the National Science and Technology progress of the second Prize 4, the National Technological Invention Second Prize 1, Provincial and ministerial awards more than 10. At home and abroad academic journals, international conferences published more than 800 academic papers, the invention of patents authorized more than 40 items. The joint laboratory currently has strong research strength in face detection and recognition, digital audio and video codec technology, content-based multimedia data retrieval, multi-function sensing technology, especially in the field of digital audio and video codec technology, face detection and recognition technology, Chinese Sign language recognition and synthesis and other fields to make many original work. My research direction and development history Under the VIPL Research Group and Joint Laboratory Research Framework, I have been focusing on computer vision (CV), pattern recognition (PR), machine learning (ML) and intelligent human-computer interface (HCI) for many years. From the perspective of computer history, "Computer Man" pursues "three high" namely: high performance, high reliability and high availability. And my research focuses more on "high availability," which is to make computers smarter and better. Better use of the computer of course, the best can be like a person can listen, will see, and even with a confidant as good understanding, anxious people anxious. The above is a broad research goal, specifically to my personal, I want to be able to give the computer like human "visual" ability, that is, the computer has not only the eyes (camera), as well as the visual cortex in the human brain, the same visual processing ability: Can judge the "see" image of the space between the eyes of the distance , what person or object is in the image, in what kind of scene, what is happening. These features are easy for people and "hard for the Sky" for computers-at least for now. Therefore, the above research goal is also a very challenging and very interesting scientific ideal. For the past more than 10 years, I have only been able to practice several of these research directions. Specifically, I started from the 1997 undergraduate graduation design, began to adhere to the present research direction: face recognition. The so-called face recognition, that is: to determine whether the input digital image or video in the face, if there is, where, there are a few, respectively, whose face. The problem seems simple and difficult to do. The international scientific community has been studying since the 60 's, and in the 70 there has been a doctoral dissertation devoted to this issue, but it has not been solved very well. By now I have studied this problem for 15 years and made some progress, for example, the performance of our algorithm has reported the best performance on many public face banks, and obtained the first place in a domestic and an international face recognition evaluation respectively with 2004 and 2006. However, there are still many unresolved challenges that plague us, making face recognition not yet fully commercialized. See here, some students may feel strange: teacher Shan face recognition for so many years, will not do something else? The problem is profound. First of all, face recognition is not a separate research topic, but a research group, including: 1) face detection, that is: Judge the image or video is not someone's face, if there are, there are several, where respectively; 2) facial feature extraction, namely: locating the position and shape of the main organ on the human face ; 3) Human face tracking in the video, that is, on the basis of detection, the human face is quickly positioned by frame; 4) face recognition, namely: Judge who's face; 5) face confirmation, that is: Determine whether the captured face is the user claims that the person; 6) face attribute analysis, that is, to determine the input face of the age, gender, ethnicity and other attributes ; 7) Facial expression recognition, that is: recognition of the face of the input faces, so as to infer their emotional state; 8) The non-photorealistic rendering of the face, that is: the input of the face of the image is automatically converted to cartoon, exaggerated comic, or modified as a painter to draw a sketch or oil-like effect; 9) face animation, Namely: direct drive input face image "Live" up, can speak, can do expression. Or it can make its cartoon image Vivid "live" up. It is not difficult for a careful classmate to think of the broad and interesting areas of application that these research directions have. Secondly, our research direction has been gradually extended from face recognition to pedestrian detection and tracking, vehicle detection and tracking, and is expanding to the next generation of intelligent video surveillance of motion recognition, behavioral analysis and abnormal event recognition up. This research direction has the important research value in the video surveillance. It is not difficult to imagine if the computer can automatically judge whether there is abnormal behavior in the monitoring video or how great the application value of the event! Again, face recognition in my case, is a research topic, more is a study of computer vision, pattern recognition and machine learning in the basic issues of the "Application case", as we do the simulation problem to master the basic principles, in my partner and I face recognition process, We are in fact more through the face recognition of the problem of "generalization" to explore these areas of the fundamental frontier topics, such as: The Theory and methods of visual modeling, visual characteristics of the design, identification features extraction, data reduction, non-linear data analysis, classifier design and so on. In fact, many of our research papers over the last few years are not related to face recognition, but are fundamental methods. Finally, I would like to emphasize that scientific research needs to accumulate and cannot be achieved overnight. Rome, Cold day. Doing research is the same, no long-term accumulation, it is impossible to create a good researcher, it is impossible to create a good research team. For face recognition such a challenging research topic, there is no long-term accumulation is difficult to make a truly innovative work, I believe that we will have experience in the future. Even after more than 10 years of research, we still have some deviations and even errors in our understanding of many of these issues-and as far as I know, this is true of many researchers, even if he is the so-called "Daniel" in this field. Our scientific research philosophy and pursuit I and our team are committed to doing international first-class (application) basic research! This sentence sounds very ordinary, but it is my personal real scientific research ideals and pursuit. I would like to take this opportunity to talk to you about the level of "international first Class" I understand. First, international first-class means always remember to be the real "wall-breaker". "Wall" refers to the boundary between the so-called "known World" and the "Unknown World" in epistemology, and I think that the real scientific research should be the process of breaking this boundary, or moving the boundary from the known world to the unknown world. Imagine a scene where people all over the world are still on this side of the wall, and only you have broken into the walls and entered the unknown realms of the people before, what a wonderful sight, what an exciting moment! The greatest pleasure in doing research is here. Of course, the walls we're going to break are definitely different from the walls that the master scientists such as Newton and Einstein have broken. The walls we're trying to break are mainly the basic problems in the fields of computer vision and pattern recognition: How to empower the computer to be as powerful as the human visual system? In fact, it is not as easy as it is to be a "real" wall breaker. In fact, many researchers are breaking the "fake walls". The reason is because it does not really deep understanding of the domain of the state of the art--this term you should often encounter, its meaning according to English is not difficult to understand, refers to a field of "the current state" or "the highest level", that is, the so-called "wall." Many researchers will not be aware of the state of the art in this area due to insufficient research or other reasons, and will do a lot of self-righteous "broken Walls" behavior: think of solving a problem, proposing new methods or achieving the best performance. As everyone knows, these problems, methods or results may already be known as the "known world" for the field counterparts, or have been in some areas of peer first step to break the wall. So, to put it simply, truly international first-class is like hundred meters, need to constantly break the existing world record (the state of the art), create our new record!        Second, the international first-class external performance is to continue to publish high-level academic papers in the top international journals and top international conferences in the field. Broken wall into the you will not automatically be recognized by other peers, the reason is very simple, others do not know that you have to enjoy the scenery in it! Therefore, must be in the appropriate situation, in a suitable way to your broken wall into the past and the view of the scenery to report to international counterparts. The most important occasion and way is to publish papers, especially in the top international journals and top international conferences of the field. Why stress is "top" and "international"? For example: A company developed a new product, feel particularly good, to put product advertising, choose a county in a province to advertise a local TV station and choose to advertise on CCTV, its influence is the same? Obviously there will be a difference! Scientific community in order to maintain scientific progress, after hundreds of years of groping, gradually established and perfected a set of academic Exchange system. The most important thing is to be a single-blind and even double-blind peer review system supported by the academic conference and the publication System of periodical papers. In order to give researchers the opportunity to publish papers, some professional academic publishing institutions to undertake a number of academic conferences and periodicals, its influence and participation is very different, even can be said mixed. So, there is the so-called "top-notch" statement, generally speaking, we only focus on the best international conferences in the field and the best 5~10 international periodicals in the field, popularly said is the cattle and cattle periodicals, the reason is very simple, these conferences and journal papers are the most noteworthy, and will be the most people read and attention. specifically to the CV/PR areas we are involved in, we believe that the most bullish will be ICCV, CVPR, nips and ECCV, while the most bull international periodicals include IEEE Trans. On Pattern Analysis and machine Intelligence (t PAMI), international Journal of Computer Vision (IJCV), IEEE Trans on Image processing (T IP), Pattern recognition (PR), etc. These meetings and some of the periodicals about 5-10 years ago, the number of articles published annually by our mainland Chinese scholars (excluding the Institute of foreign companies) is almost single-digit. In recent years, more and more, of which our group has become an important force, since 2005, our research group has published more than 30 articles in these conferences and periodicals, thereby establishing a good image in the international. Such a good article publication record should also be our group graduate doctor can find a good position of important reason, so-called a successful team did not fail the individual! Finally, and most importantly, international first-class means the need to pursue the real impact of scientific research work (impact). The number of "top conferences and periodicals" mentioned above is important, but the number is not an essential factor. For us, the "top conferences and periodicals" paper, from scratch, from little to many, is indeed a great improvement, but what we are pursuing is "substantial influence". The most important measure of real influence is the number of times the paper is cited by the international counterparts, which can be said to a large extent reflects the weight of the paper in the eyes of peers. Imagine that you published a paper 10 years later by the international peer cited thousands of times, that is how proud of a thing Ah! And, generally speaking, a paper with a high citation rate will often become the content of a classic textbook several years later, which is even more commendable, because it means that your paper has truly become a knowledge point in the "treasure trove of human knowledge" and is well known and even adopted and used by more people. Of course, the paper is quoted need time accumulation, not at the time of the publication can be judged, but according to the general law, in the field of top-level conferences and periodicals will have more people to pay attention to, usually have a higher citation rate, this is not difficult to understand. A large number of papers published in the three-stream conferences and periodicals are often cited in very few cases, and even a large percentage is never quoted, and such articles are called "Junk papers". Spam articles are a matter of no benefit to a truly serious researcher and its research team, and will greatly affect its academic prestige. If the citation rate represents the academic value and theoretical significance of scientific research work, practicality is another important evaluation element of the research achievement. This is especially important in the field of computer science and technology, because computers are still engineering in nature and must emphasize their application value to see if they really have actual or potential application needs. At least our team doesn't appreciate doing purely mathematical or word games in the field of computer technology. So, over the past few years, we have partnered with nearly 10 international companies (including NEC, Panasonic, Nokia, Samsung, Qualcomm, Microsoft and Huawei, Baidu and Silver Chen Technology) to successfully export many of our technologies to the enterprise for application. In short, to do the international first-class application of basic research is our consistent pursuit of this team, so we have done our best to create the academic atmosphere that can support such lofty pursuit, in the daily management of graduate students, we always preach such lofty ideas. What kind of people are not fit to join this research team I would like to contact my graduate students to take a serious look at the question: "Am I suitable to join the Chinese Academy of Sciences Computational visual information processing and learning research group?" ”。 To help you think about this, I first tell you what kind of classmates are not fit to join our team: (1) No interest in scientific research, computer vision and pattern recognition in the field of students who have no enthusiasm. Interest is one of the sources of scientific research and the best teacher for your progress. So, with a strong interest and enduring passion, you're half done. On the contrary, no interest, no enthusiasm means that you will spend a few years in the "pain" good time, it is not worth the candle, so, please do not apply himself! (2) students who are purely for diplomas and degrees. I must state that the Boljan of our group is generally 6 years old, of course, this does not mean that you must be 6 years to graduate, as long as you work hard enough to achieve the laboratory identified research level (and you are willing to graduate early), 5 years or even shorter also can graduate, such examples are also many. But if you're purely for a diploma, you can choose a mentor and a team that will be able to graduate for 5 years or less. (3) students who do not want to continue to engage in scientific research after graduation should be careful to apply for this task group. We believe that we are mainly trained in research-oriented talents. If you do not do the relevant research work after graduation, it is not only a waste, but also will give you later to engage in other specific types of work to bring some unfavorable factors. For example, if you want to do business after graduation, I suggest you should not read a doctor. This, of course, varies from person to person. (4) students who are very afraid of maths. The research direction of this group involves more mathematics knowledge (especially high number, line generation, probability, statistic theory ...). If you do not have a solid mathematical foundation, it means that you lack the ability to formalize and translate visual problems into known mathematical models and then solve them, which means it is difficult to innovate. This is not to say that you must be a math expert, or that you can do well in math. What we need more is the ability to transform scientific problems into abstract mathematical expressions and solutions. (5) students who have no confidence in their English reading and writing ability in both English and Chinese. The importance of expression in both English and Chinese is self-evident: any scientific research must be presented in the appropriate language, and a top international journal paper usually has twenty or thirty pages of English, and no good writing ability is difficult to do this task. (6) students who are not mature enough to think. For example, think that the teacher is the boss of the exploitation of their own labor and thus negative to do research, do not like independent thinking only know the passive waiting for teachers to decorate homework, 60 points long live graduation can students. In addition, the most important point, do not want to know what they really want to students, it is best not to join our research team. In short, I would like to ask you to consider a question: Am I really interested in doing research? Do you really want to do the research? The reason I emphasize " really " is because, as far as I know, because you are not fully mature yet, most of the students do not want to know what shouting truth is interested in, nor do they want to know what it means to do research. Please think about what will happen in the 6 years or so of your PhD. Your undergraduate or even graduate students graduated several years later, may already have a good material life, began to wear brand-name clothes, with brand-name products, and even began to buy a car to buy a house marriage and children. But you are still "studying", from the family, friends, the reality of pressure gradually increased. So in this case, are you still willing to stick to your own interests? Insist on doing research? That's the point. The so-called "gain must be lost", in order to study, for the sake of interest, it means that we must abandon something, or at least temporarily abandon something. Would you like to give it up for the moment? If not, then I am afraid it is not really interested, not really want to do research. If so, then you must choose the other teachers as early as possible-or even if you are admitted, also means that you may be in the best years of life, spent a "very unhappy 5 years", it is not worth it!! What will you gain by joining this research team? I hope to recruit graduate students can become my "like -minded win cooperation ." The so-called "like-minded", is at least in the next 5-6 years unswervingly appreciate and willing to practice what I said before the scientific research concept and pursuit, willing under my guidance, together to do international first-class research. In order to explain the so-called "win", I would like to enumerate here if you do research with heart what kind of harvest. (1) You will gradually grow into a mature researcher in the field of CV/PR. This will provide a good starting point for your future research in related fields. This includes a well-documented list of published papers, as well as good research habits, time concepts, scientific and technical paper writing norms and techniques, good academic ethics, but also includes the truth of doing things and learning. In particular, your efforts to publish top international conferences and journal papers will give you a high starting point in this area. (2) You will learn the "Tao" of being a man and doing scholarship. Scholarship is important, the truth of doing things is equally important. In our case, you will learn, realize and practice the truth of doing things and doing scholarship. (3) You will be given the appropriate "funding" to complete your studies. In particular, we need to tell you that: all the training costs of graduate students in CAS are to be paid from the teacher's project funds. In other words, the tutor needs to pay for your tuition, accommodation, postgraduate benefits, hardware equipment, station rental fees, RA allowances, water, electricity, network, printing supplies, paper publishing fees, participation in international conference registration fees and travel expenses, and so on. On average, each graduate needs at least $40,000 a year, which means that a master's 3 years will cost about $120,000, while a PhD student will cost at least $200,000 in 5 years. All these expenses are to be paid for by the tutor from the project funds! (4) You will receive a postgraduate allowance to maintain your basic life. As mentioned earlier, our doctoral students receive a monthly stipend. This includes 1500 yuan per month from the Institute of Postgraduate Subsidy (of which 500 yuan from the tutor to pay the training fee), the other team will also issue a monthly 1000~2500-yuan Ra allowance (will be adjusted according to the contribution), that is, Excellent doctoral students can receive a maximum of 3500 yuan per month. In addition, the Institute, the Chinese Academy of Sciences and so on each year there are a number of scholarship awards outstanding performance of graduate students, the award amount of thousands of yuan range. In addition, each year, the laboratory will appropriately reward students who have published papers in the above-mentioned top international conferences and periodicals. A reminder: If you want to buy a house through these subsidies, it's definitely unrealistic! ：） (5) You will be given the opportunity to attend international conferences or to communicate abroad. In general, we will arrange for doctoral students with top international conference articles to participate in international conferences in order to reach out to the important international counterparts in the field, and to engage in a face-to-head communication with them in order to broaden our horizons. In recent years, with the increase in cooperation with international counterparts, more and more doctoral students have been to study abroad (usually six months to a year) opportunities, we are also trying to create more such opportunities. (6) You will end up with a truly coveted doctorate. Our goal is to develop a PhD that can be competitive with graduates from prestigious universities in the United States. I personally think that the average doctoral level we are currently cultivating is comparable to that of the University of America's computer department at around 20-30 universities, some of which have been able to compete in some ways with those of the top universities in the United States, such as in the publication of articles. Through our efforts, I believe we can do a better job and end up being on par with the top 10-20 of American universities. (7) You will get a good referral job opportunity. We have close contact with well-known international counterparts in many fields, and even project cooperation, which means that we have the opportunity to recommend excellent graduate students to enter their research group, and even many of them will take the initiative to contact us, ask us to recommend excellent graduates to them. About interviews and competitive situations Many students in contact with me, will ask: "Teacher you think my situation is there any hope that you will be accepted?" The answers to this question are influenced by several factors. First, about the degree of competition. As for the experience of the past few years, I have been contacted by many students each year. Take 2009 years as an example, I have a general statistics, and I contacted the students have more than 50 people, of course, this does not mean that all these students will eventually in the network registration system to report me, but almost all will be about 20 people to fill me. So, overall, the competition is more intense. The good news is: first, our subject group of teachers will be unified every year to arrange enrollment (unified interview and scoring, finally in accordance with the overall ranking admission), so, if the other teacher's classmates less, we will be in the group of several mentors to adjust, so that the overall admission rate will be higher; second, many alumni because " Graduation time is long, can not agree with my idea or fear of competition and challenge "and choose to quit." Second, about what qualities I value most. What I value most is whether the candidate really agrees with my point of view, especially if you are really interested in research, visual pattern analysis and recognition, whether you are really willing to use your 12 times times enthusiasm to study it without regard to other temporary gains and losses. Of course, Mathematics (especially advanced mathematics, linear algebra, probability theory and statistics), Chinese and English expression and programming proficiency is my most concerned about the basic ability. Third, about rankings. You do not have to be within the first 1%, but within 10% is necessary, preferably within 5%. Four, about contests and project experiences. To me personally, there is no uniform, not a necessary condition. But if there is, I would be interested to know what you have done in it. Five, about the interview. Over the years, our group is the implementation of collective interviewing system, that is, by the 5 teachers to conduct a collective interview scoring, and combined with the written test and machine test results , do the final total ranking. The usual process of interview is first by the candidate to do 5 minutes of self-introduction, please prepare PPT to our Admissions Secretary Wang Xiaobiao teacher ([email protected]), remind everyone in the PPT focus on their highlights, and please do not time out. Otherwise, the interview time (usually about 15 minutes) will be compressed. Six, about the written test. The written test is usually 2-3 hours, mainly on the basis of mathematics, including high-number, probability, statistics, linear algebra and other mathematical courses, as well as Chinese and English expressions (such as writing or wrong). Seven, about the machine test. Machine test time 1 hours, usually has 5-10 questions, need to be written and debugged in Visual C + + environment, and be able to output the correct results for the test sample. The VIII, about pre-screening. Due to the large number of applicants, we will generally according to the information provided by the situation, the students enrolled in pre-screening, through pre-screened students can participate in the interview. We will send email and other means to inform you whether the pre-screening. Finally, I wish everyone success! Shanshi The draft was revised in September 2009, August 23, 2012.

From:http://www.jdl.ac.cn/user/sgshan/tostudents.htm

To contact the students of my exam-exempt graduate student-Shanshi researcher-Institute of Computational Sciences