Author: Jun Dai

Last update: 2025.06.29

1. Pre-word:

People pursue research for many different reasons — curiosity, career advancement, financial gain, or other motivations. Regardless of their driving force, outstanding researchers always have a clear sense of where they are and what their next steps should be. I've heard many people say "you should research what you truly love," and I completely agree! That's why I've gathered and organized the research directions that fascinate me into my own integrated Research Map. After reading this, I hope you'll say, "This research is truly needed by the community, and the techniques developed are meaningful and innovative!".

                                                     Fig 1. Which kind of paper you want to finish?

The basical learning style of me:

                                                      Fig 2. Treat me as a perception machine!

1. My gold-standards for great research:

I'd like to share what I consider to be great research works and papers.

• Works: should be viewed at a system level, meaning they comprise multiple papers published across top venues like CVPR/ICCV/ECCV/SIGGRAPH. They are systematic projects that either solve or define specific problems — for example, how SAM revolutionized segmentation tasks, or how Google's HDR+ demonstrated the possibilities of mobile photography.
• Papers: should be evaluated according to fixed standards and viewed at element-level, such as the importance of the problem being addressed and whether the experiments adequately validate the authors' claims. I really like the analogy from Bill Freeman about which kinds of papers are more likely to be accepted by top conferences;

Common standards for Works and Papers:

• Should be meaningful and well-motivated;
• Should be make sense and full of logics;
• Should be elegant and more simple more better;