Presentation on theme: "How to finish your project on time ?"— Presentation transcript:
1 How to finish your project on time ? A.K.A. How to graduate on time ?
2 Know the graduation requirement Which SCI journal ? / Conference ?CVPR/ICCV or ACCV/ICIPPAMI/IJCV or CVIU/CVATypically, graduation requirement decrease if you spend more time in graduate school…3rd year – a lot of good publications4th year – good publications5th year – publications6th year – publications…
3 Timeline for a PhD degree (4th years) 1st year – 2nd year: taking courses3rd year - ?? Year: working on research projectKeep in mind the timeline of review processConference usually takes half year for review processJournal can sometimes take 1- 2 years…Before your graduation, you need to spend time to write your thesisIt can takes your several months to half yearKeep in mind that you also need to do job hunting before graduation
4 Why some people can have a lot of publications before their graduation ? Self-motivated / Hard working ?i.e. no sleeping, no entertainmentThis is one reason, but not the main reasonLuck ?i.e. Lucky to have “good” reviewers who pass your submissionThe is also an important reason, but first, your submission need to meet the bottom line requirementFriends / Co-authors?Ask your friends to help for your experimentsIncrease your publications with many co-authorshipBut, you still need to have at least 1 paper which you are the first authorWork smart !This is the most important reason
5 Work hard ≠Work smart Don’t waste your time, spend your time wisely ! Some students spent a week finishing a research project while some other students spent a whole year but he/she still cannot finishWhy ?Some students spent a year working on a research project, and then, he/she found out her project is unsolvable!Except crying, what else…?Whose responsibility ? Student / Advisor ?Who suffered ? Student / Advisor ?
6 Before you jump into a research project Your advisor ask you to do a research project does not means that you have to do itOf course, you will make your advisor unhappy at that momentBut, your advisor’s unhappy last for only several minutes, but your unhappy last for several months/yearIf possible, ask more details about that projects, as much as you canUnderstand the nature of the project, is it just a brainstorm project? or is it a project with long term research plan?Hint: Never say “yes” immediately, say “I am going to think about it, and I will get back to you on Friday / in a week”Now, you buy yourself more time to think about the project…
7 What are you going to do in that week ? Sleeping / Playing / Dating ?Don’t be stupid, you buy your time by confronting with your advisor and then you waste itThinking about my own schedule, e.g. you might have freelance / travel plans / other worksThis is a consideration, but this is not the most important thing you need to considerAfter you say “no”, your advisor would have other tasks to you. Do not think you are going to be free after saying “no”What should you do ? After one week, you have to say “yes/no”
8 During that week1. Do a quick survey about the state-of-the-art techniques !Collect most relevant papers, read them and understand themThis has to be done in your first three days (Assuming you have 7 days)Keep in mind, if you lose this golden period to say “no”, you are going to suffer for a whole yearIf you say “no” after several months, your advisor would think you did nothing in the last several monthsIf your advisor has some resources to you, read them and understand all of them
9 During that weekAfter you do your survey, spend time to think about the problem, the input, the output and the project goalTry to write down your solution / planningDo a simple experiment to test the feasibility of the projectIf possible, implement one / two previous worksBe open minded, if the direction proposed by your advisor is a dead corner, try to change the direction without offense to your advisorAt the end, your advisor just want to have a publicationThis is a busy week, but again, if you waste this week, you will even waste more time in your future
10 After that week Give a presentation to your advisor Describe the survey you have doneSummarize the pros/cons of previous worksDescribe your understanding about the projectYour understanding might not be the same as your advisor. Now, it is the good time to come up with a consensusIf you say “yes”, describe your solutionas details as possible, it should go to the very low level as much as you canIf you say “no”, describe your reasonyour reason can be:This direction has been done, cite the paper, and refers to the survey you have doneThis direction is infeasible, if you have done simple testing, show your resultsIf you found something interesting, but different from the direction proposed by your advisorThis is also a golden opportunity to propose your own ideaIf your advisor like it, you can work on it, else, choose other topic
11 Keep in mindThis one week is a golden week, a week that determine your works for the future several months or yearsIt is difficult to say “no” again after this weekIf the direction is wrong, you are going to waste a lot more time than a weekAgain, if you work on it for a year and then you fail, you are going to spend one more year for your graduationIt might be good, because the graduation requirement is lower…
12 Now, what should you do after you start the project Be self-motivatedKeep in mind, this is now your direction, you choose this direction yourself, and be responsible to yourselfArrange your time wiselyAgain, you need to arrange your time wiselyBut how ?
13 Understand your project Divide and conquerDivide your project into a lot of sub-componentsSome components are important, some are notIdentify which component is the most important, and spend your time depending on the importance of each componentBe smart, some components can be “Hack”, e.g. user inputs/correctionsKnow how to skip the unimportant taskIf there are source codes available, use them. Don’t be stupid to implement everything yourselfIf your friends have experiences on some sub-tasks, go ask themDon’t get stuck in some unimportant task
14 When doing your experiments… Pick the examples wiselyStart from easy example, after successful on one example, try a little bit more difficult oneDo not pick the examples that surely violate your assumptionTesting on environment variables / parametersFix the variables/parameters as much as you can, each time, only change the value of one parameters.Some parameters are critical, some are not, you will find our after this experimentIf you have many parameters, write a script to test all parameters automaticallyDo not sit in front of your compute to test your parametersParameter tuning can be the most wasting time process, do it wisely