Contents
1 Introduction 1
2 Executables 2
3 Building Stasm 4
4 Stasm library functions 5
4.1 Simple interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
4.2 More versatile interface . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
4.3 The minwidth argument . . . . . . . . . . . . . . . . . . . . . . . . . . 5
4.4 The multiface argument . . . . . . . . . . . . . . . . . . . . . . . . . 5
4.5 Partial initialization by the user . . . . . . . . . . . . . . . . . . . . . . 6
4.6 Error handling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
4.7 Utility functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
5 Visual FAQ 8
6 Stasm version 4 and 3 11
7 Notes on the implementation 12
8 Citing Stasm 14
9 Acknowledgments 15
A Stasm License Agreement 17
Bibliography 19
1 Introduction
Stasm is a C++ software package for locating facial landmarks in faces. You give it a
face image and it returns the positions of the landmarks (Figure 1).
Stasm is designed to work on front views of more-or-less upright faces with neutral
expressions. You will see poor fits on faces at angles or with expressions. Download the
self-installing Windows program wasm.exe to see what Stasm can do. It has performed
well in comparative studies [3].
Stasm uses the OpenCV library and is released under a BSD style license (Appendix A).
The Histogram Array Transform (HAT) descriptors used by Stasm for template match-
ing are similar to SIFT descriptors [9]. Since the SIFT algorithm is patented, this
potentially opens a patent issue for commercial use. (I’m not qualified to say if it
does.) Section 7 has some technical details.
If you are interested in training new models, please see Building Stasm 4 Models [11],
downloadable from the Stasm website.
1
评论0
最新资源