15
Paper Gestalt Carven von Bearnensquash

Paper Gestalt Carven von Bearnensquash. Background Peer review imperfect review process Growth in the volume of submissions, tripled over the last 10

Embed Size (px)

Citation preview

Page 1: Paper Gestalt Carven von Bearnensquash. Background Peer review  imperfect review process Growth in the volume of submissions, tripled over the last 10

Paper Gestalt

Carven von Bearnensquash

Page 2: Paper Gestalt Carven von Bearnensquash. Background Peer review  imperfect review process Growth in the volume of submissions, tripled over the last 10

Background

• Peer review imperfect review process• Growth in the volume of submissions, tripled

over the last 10 years• Less than ideal pool of reviewers• General layout of a paper

Page 3: Paper Gestalt Carven von Bearnensquash. Background Peer review  imperfect review process Growth in the volume of submissions, tripled over the last 10

Abstract

• Intuition: Quality of paper general layout of the paper

• Computer vision techniques to predict if the paper should be accepted

• Result: reject 15% of good papers, cut down the number of “bad papers” by more than 50%

Page 4: Paper Gestalt Carven von Bearnensquash. Background Peer review  imperfect review process Growth in the volume of submissions, tripled over the last 10

Related work

• Unique work• Text based – biased to certain terms:

“boosting”, “svm”, “crf”, ignores rich visual information

• No previous work known

Page 5: Paper Gestalt Carven von Bearnensquash. Background Peer review  imperfect review process Growth in the volume of submissions, tripled over the last 10

Approach

• Formulated as a binary classification task• Training data set of example-label pairs, {(x1;

y1); (x2; y2); ...(xn; yn)}, Xi: feature values for paper i, Yi: binary label, “good” or “bad”

• Goal: learn a function f: X {0, 1}

Page 6: Paper Gestalt Carven von Bearnensquash. Background Peer review  imperfect review process Growth in the volume of submissions, tripled over the last 10

Approach

• Adaboost

• Select feature classifierwith lowest error rate, increase weight of mis-classified data

Page 7: Paper Gestalt Carven von Bearnensquash. Background Peer review  imperfect review process Growth in the volume of submissions, tripled over the last 10

Approach

• Empirical error is bounded by

• More math: Include Maxwell’s equations in the paper

• Equations improvepaper gestalt

Page 8: Paper Gestalt Carven von Bearnensquash. Background Peer review  imperfect review process Growth in the volume of submissions, tripled over the last 10

Features

• gradient, texture, color and spatial information

• LUV histograms, Histograms of Oriented Gradients and gradient magnitude.

Page 9: Paper Gestalt Carven von Bearnensquash. Background Peer review  imperfect review process Growth in the volume of submissions, tripled over the last 10

Experiments - Data Acquisition

• Accepted papers from CVPR 2008, ICCV 2009, and CVPR 2009 as positive examples #1196

• Workshop papers from these same conferences as an approximation as negative examples #665

• Papers converted to images, resized and padded with blank pages.

• 25% testing and 75% training

Page 10: Paper Gestalt Carven von Bearnensquash. Background Peer review  imperfect review process Growth in the volume of submissions, tripled over the last 10

Experiments -

• Assuming that rejecting 15% of good papers is acceptable, we can cut bad papers in half

Page 11: Paper Gestalt Carven von Bearnensquash. Background Peer review  imperfect review process Growth in the volume of submissions, tripled over the last 10

Experiments

• “we’re not sure what this figure reveals”• bar plots are particularly aesthetically pleasing

Page 12: Paper Gestalt Carven von Bearnensquash. Background Peer review  imperfect review process Growth in the volume of submissions, tripled over the last 10

Experiments – good examples

Page 13: Paper Gestalt Carven von Bearnensquash. Background Peer review  imperfect review process Growth in the volume of submissions, tripled over the last 10

Experiments – bad examples

Page 14: Paper Gestalt Carven von Bearnensquash. Background Peer review  imperfect review process Growth in the volume of submissions, tripled over the last 10

Experiments – the paper itself

• The system reported a posterior probability of 88.4%, which reassured us that this paper is fit for the CVPR conference.

Page 15: Paper Gestalt Carven von Bearnensquash. Background Peer review  imperfect review process Growth in the volume of submissions, tripled over the last 10

Conclusions

• The quality of a computer vision paper can be estimated well by basic visual features

• A real-time system to predict weather a paper should be accepted or rejected