To create your own community (CVPR exhaustive survey)
Do you want to have your own community? I have led a research community cvpaper.challenge since May, 2015. We are treating paper survey and conducting academic research. Now in 2021, over 500 people belong to the research community. At the beginning, we were 5 people including 3 undergraduate students, a graduate student, and me (I was a post-doc researcher). The initial time was a very small-sized team.
On one hand, sometimes we count several important elements to build a good team. For examples, (i) human resource, (ii) facility, and (iii) information are one of them. In my case, I didn’t enoughly have (i) and (ii) because undergraduate/graduate students were beginners in the field and in those days, it was very early career as a researcher for me. Therefore, I must dramatically increase (iii) information for expanding the research community. Fortunately, at that time (of course at this moment!), we could collect high-quality information from open access sites e.g., arXiv.com and Computer Vision Foundation (CVF). I heard that some top researchers look through all the papers in the top conferences. In case of computer vision field, it is said that there exists three top conferences, namely CVPR, ICCV, and ECCV. Therefore, we decided to comprehensively read one of the top conferences in a short term. The upcoming top conference was CVPR 2015 which is held in June, 2015. In this way, as the first trial of research community, we decided to conduct an exhaustive CVPR survey.
As you imagine, the CVPR survey for 602 papers by 5 people, 4 beginners out of 5, was extremely hard! We all are *NOT* native English speakers. At the beginning, an undergraduate student spent for a week to read a paper in CVPR. I also devoted a couple of hours in a paper. We were in a desperate situation… Therefore, I re-defined the rule of challenge as (1) one-page summary creation, (2) we must read at least abstract and introduction. We shared the work space on Google Drive doc (as the bottom figure), and paper list to grasp unread papers.
We have translated CVPR papers from English to Japanese for sharing Japanese community. At the same time, we tweeted on our account (as the bottom figure). This generated a synergy effect for spreading our research community with the crazy idea! We declared that I would completely read the 602 CVPR’15 papers and create their summaries for the next months.
Basically, a researcher can read a paper want to read. There is no #paper quota. However, every weekend a researcher must report the number even though the #paper is zero. Also, we had discussed how to effectively read a paper every week. We had gradually increased the #paper, month after month as follow.
Finally, in the four months later, we achieved to complete the exhaustive CVPR survey! Through the survey, we got the following benefits.
- We simply enjoyed reading the CVPR papers!
- Undergraduate/graduate students were no longer resistant to English papers. They have read around 100 papers through the challenge.
- Undergraduate/graduate students got a great motivation to their research for a conference paper.
- We acquired a lot of perspectives for creating research themes.
- We noticed that the most important thing is ‘how to set/create a research question’.
Then, we received an email to have an invited talk about the challenge. We also decided to summarize the survey paper for CVPR 2015 and the challenge. We made the talk slide in Japanese and arXiv paper in English. We are happy to share the contents. The talk slide is over 10k views and the arXiv paper was weekly 1st place on Arxiv Sanity Preserver.
