How to PhD-research (10 steps)


  1. Select topics – preferably choose emerging and hot areas unstudied by others.  Talk to your advisor for help. For computer vision, see these lists:

Keith Price Bibliography Annotated Computer Vision Bibliography


  1. Determine what you want to do for the selected topic.  State the problem in general terms.


  1. Thoroughly collect materials of previous work – papers, projects, etc usually by web search, paper search, or person-lab search.


  1. Be prepared to understand those materials – mathematics, notation, statistics, key ideas.  If you are not ready, read relevant textbooks and study relevant materials.  Talk to area specialists for help.  Even talk directly to paper authors for help.


  1. Be sure to implement their ideas by yourself.  You may talk to the authors for help.


  1. Specify your problem in details.  What is your specific goal?  What is your approach compared with others?  What are your contributions?


  1. Think and think – improvement, apply different methods, create your own algorithm, develop new techniques, and create new concepts.


  1. Implement your idea and compare it with theirs.  Review.


  1. Make reports and demo and publish them.


  1.  Discuss topics with colleagues and make connection with them.



Kyungnam Kim


Computer Science Department

University of Maryland, College Park