🚀 go-pugleaf

  [Dr. David Kahaner is a numerical analyst on sabbatical to the 
   Office of Naval Research-Asia (ONR Asia) in Tokyo from NIST.  The 
   following is the professional opinion of David Kahaner and in no 
   way has the blessing of the US Government or any agency of it.  All 
   information is dated and of limited life time.  This disclaimer should 
   be noted on ANY attribution.]

  [Copies of previous reports written by Kahaner can be obtained using
   anonymous FTP from host cs.arizona.edu, directory japan/kahaner.reports.]

To: Distribution
From: 
 David K. Kahaner
 US Office of Naval Research Asia
 (From outside US):  23-17, 7-chome, Roppongi, Minato-ku, Tokyo 106 Japan
 (From within  US):  Unit 45002, APO AP 96337-0007
  Tel: +81 3 3401-8924, Fax: +81 3 3403-9670
  Email: kahaner@cs.titech.ac.jp
Re: Facial classification by neural nets
25 Dec 1992
This file is named "face-nn.92"

ABSTRACT. Report on an experiment using a neural network for sex
descrimination based on facial classification, from Osaka University.

[ Quoted from JETRO, 11/92, p25]

Man-and-Woman Classification from Human Face by Neural Network

Prof. S. Tamura at Osaka University Medical School, and Asst. Prof.  H.
Kawai at Osaka Electro-Communication Junior College have demonstrated
that the back propagation neural network can identify with high accuracy
the sex of persons from their faces.

	Prof S. Tamura 
        Osaka University
        Medical School
        2-2, Yamadaoka, Suita City, Osaka 565 JAPAN
         Tel: +81-6-875-7111; Fax: +81-6-875-7352

The demonstration test was made with trimmed photos of faces.  The
photos had the skirts cut away to exclude hair style and contour, which
are likely to be different between men and women.  Each photos was
transformed into three mosaic images: arrays of 8x8, 16x16, and 32x32
squares, with every square re-painted with gray at the mean brightness
of the square.

The test started by preparing mosaic facial images from 30 persons, half
male and half female.  The neural network learnt the images and sex of
the subject.  The test encompassed images of all three square arrays,
but learning at one time used a series of 30 images of one array.

Next, face photos of 30 other persons consisting of 15 men and 15 women
were transformed into mosaic images of the three square arrays.  The
neural network was given the new images of one array, and checked to see
correct identification of sexes.  When the neural network learned
32x32-array mosaic images, it answered 93% correctly even for given
8x8-array mosaic images, which were not recognizable as human faces to
people.

When people identify the sex of a person, they mostly look at the
behavior, clothes, hair style, and the like.  People may identify the
sex by only looking at a person's face, but here the identification
heavily depends on the contour of the face.  Even clipped photos used to
make the mosaic transformation in the test often make it difficult for
people to make the sex identification.

The test indicated that people have an identifying ability poorer than
the neural network [sic].  How the neural network differentiates between
the sexes has to be elucidated.  The neural network may extract facial
features that people cannot detect.


----------------------------END OF REPORT---------------------------------