Best Papers
CSC 262 - Computer Vision - Weinman
1 Introduction
We will read and discuss in class one or two of the best papers from
the most recent top systems conferences. In this way, we'll be learning
together:
- How to read research papers
- About the latest in computer vision research
- What the community thinks is currently important
2 Candidates
Our candidates (listed in no particular order) are drawn from CVPR
2021, CVPR 2022, ECCV 2022, and ICCV 2021. See the list of papers
below and read their abstracts.
- Task
Programming: Learning Data Efficient Behavior Representations. Jennifer
J. Sun, Ann Kennedy, Eric Zhan, David J. Anderson, Yisong Yue, Pietro
Perona. (CVPR '21).
- GIRAFFE:
Representing Scenes as Compositional Generative Neural Feature Fields.
Michael Niemeyer, Andreas Geiger. (CVPR '21)
- Learning
to Solve Hard Minimal Problems. Petr Hruby, Timothy Duff, Anton
Leykin, and Tomas Pajdla. (CVPR '22)
- EPro-PnP:
Generalized End-to-End Probabilistic Perspective-n-Points for Monocular
Object Pose Estimation. Hansheng Chen, Pichao Wang, Fan Wang, Wei
Tian, Lu Xiong, Hao Li (CVPR '22)
- Pixel-Perfect
Structure-from-Motion with Featuremetric Refinement. Philipp Lindenberger,
Paul-Edouard Sarlin, Viktor Larsson, Marc Pollefeys. (ICCV '21)
- Swin
Transformer: Hierarchical Vision Transformer Using Shifted Windows.
Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen
Lin, Baining Guo (ICCV '21)
- On
the Versatile Uses of Partial Distance Correlation in Deep Learning.
Xingjian Zhen, Zihang Meng, Rudrasis Chakraborty, Vikas Singh (ECCV
'22)
3 Voting
Please vote by emailing your TOP TWO choices (by number) to the instructor
by Wed 30 Nov.
4 Responses
You will be required to submit a brief 225-275 word critical response
to
each paper before class to help prepare you for the
discussion. In particular, you should note:
- What problem are they trying to solve?
- Why is the problem important?
- How does it currently get done and what are the limitations?
- What are the authors' goals?
- Does the paper have a scientific thesis? Is it falsifiable?
- What are the paper's claims?
- Are the claims substantiated (by theory or experiment)? If so, how?
- What are the limitations of the proposed approach?
- Are there ways to extend the method?
You should include at least two primary points that critique, dispute,
extend, or reinforce the paper. Submit your responses (in PDF format
only) via Gradescope; they are due at the beginning of class on the
day of discussion.
Acknowledgments
The questions above are inspired by and adapted from the following
works.
Fong, Philip W.L., Reading a computer science research
paper ,
SIGCSE Bulletin 41, 2 (2009), pp. 138-140.
doi:10.1145/1595453.1595493
Keshav, S., How to read a paper ,
SIGCOMM
Computer Communication Review 37, 3 (2007), pp. 83-84.
doi:dx.doi.org/10.1145/1273445.1273458