A summary of Robust Real-Time Face Detection

P. Viola and M.J. Jones; DOI

For the summary of the paper, go to the Summary section of this article.

Table of Contents

• A summary of Robust Real-Time Face Detection
  • Table of Contents
  • First Pass
    • Category
    • Context
    • Contributions
  • Second Pass
    • Background Work
    • Motivation
    • Figures, Diagrams, Illustrations, and Graphs
    • Clarity
    • Relevant Work
    • Future Directions
  • Summary
  • Summarization Technique
  • Citations

First Pass

Discussion about the title, abstract, introduction, section and sub-section headings, and conclusion

The paper Robust Real-Time Face Detection by P. Viola and M.J. Jones [1] presents a new methodology for efficiently performing face detection. They due this through the usage of an integral image which is able to reduce the computational complexity to constant time (O(1)) of analyzing an image as it doesn’t rely on scale invariance and thus an image pyramid. Additional, the classifier that they build is “simple and efficient” and allows for the engineer to specify a large number of features to be analyzed without compromising on performance as it relies upon the Ada Boost algorithm to select important features. Furthermore, the authors propose a method for building a cascade of classifiers which further reduces computation time as each classifier specifies . Finally, they propose experiments that can be ran on face detection data sets to conduct supervised learning.

While this paper does propose many new and innovative ideas, the paper originates from 2003.

Category

What type of paper is this work?

This is a research paper focusing on improving the Computer Vision task of face detection without the reliance of CNNs.

Context

What other types of papers is the work related to?

This paper is most closely related to non-CNN face detection papers.

Contributions

What are the author’s main contributions?

They create an integral image which is able to reduce the computational complexity to constant time (O(1)) of analyzing an image as it doesn’t rely on scale invariance and thus an image pyramid. Additionally, the classifier that they build is “simple and efficient” and allows for the engineer to specify a large number of features to be analyzed without compromising on performance as it relies upon the Ada Boost algorithm to select important features. Furthermore, the authors propose a method for building a cascade of classifiers which further reduces computation time as each classifier specifies . Finally, they propose experiments that can be ran on face detection data sets to conduct supervised learning.

Second Pass

A proper read through of the paper

Background Work

What has been done prior to this paper?

Prior work has been done in creating face detection systems. Prior work has been done in creating the Ada Boost algorithm that is used to create a cascade of classifiers. Prior work has been done in identifying methodologies to create image features.

Motivation

Why should we care about this paper?

We should care about this paper as it presents a non-CNN methodology for reliably identifying faces in images. Additionally, the authors also present a methodology for doing this task efficiently on low end hardware.

Figures, Diagrams, Illustrations, and Graphs

Are the axes properly labeled? Are results shown with error bars, so that conclusions are statistically significant?

The figures, diagrams, and graphs are well explained and designed.

Clarity

Is the paper well written?

Yes, however a bit lengthy. Optimizations could have been made with respect to reducing the amount of content describing the background to the Ada Boost algorithm.

Relevant Work

Mark relevant work for review

The following relevant work can be found in the Citations section of this article.

• A Decision-Theoretic Generalization of On-Line Learning and an Application to Boosting [2]

Future Directions

My own proposed future directions for the work

I’d like to implement their work on a low powered device and compare it to a newer CNN model on ML metrics.

Summary

A summary of the paper

The paper Robust Real-Time Face Detection by P. Viola and M.J. Jones [1] presents a new methodology for efficiently performing face detection. They due this through the usage of an integral image which is able to reduce the computational complexity to constant time (O(1)) of analyzing an image as it doesn’t rely on scale invariance and thus an image pyramid. Additionally, the classifier that they build is “simple and efficient” and allows for the engineer to specify a large number of features to be analyzed without compromising on performance as it relies upon the Ada Boost algorithm to select important features. Furthermore, the authors propose a method for building a cascade of classifiers which further reduces computation time as each classifier specifies . Finally, they propose experiments that can be ran on face detection data sets to conduct supervised learning.

The main “wow” factor of this work is that it was built on a low powered system. This same application could be more performant on modern smartphones in comparison to the system that it was originally tested on.

Summarization Technique

This paper was summarized using a modified technique proposed by S. Keshav in his work How to Read a Paper [0].

Citations