Zuletzt angesehen:
Seite anzeigen
Ältere Versionen

Unterschiede

Hier werden die Unterschiede zwischen zwei Versionen angezeigt.

Link zu dieser Vergleichsansicht

Beide Seiten der vorigen RevisionVorhergehende Überarbeitung
Nächste Überarbeitung		Vorhergehende Überarbeitung
face_detection:torch3vision [2009/05/26 10:55] – geraldface_detection:torch3vision [2024/02/29 13:36] (aktuell) – Externe Bearbeitung 127.0.0.1
Zeile 2: Zeile 2:
  
 {{tag>face_detecton gesichtererkennung}} {{tag>face_detecton gesichtererkennung}}
 +
 +jpegtopnm is obsolete under debian. Use djpeg instead
 +<code>apt-get install libjpeg-progs</code>
  
 [[http://torch3vision.idiap.ch/]] [[http://torch3vision.idiap.ch/]]
  
-Quelle((http://ultrahigh.org/2008/10/14/command-line-face-detection/))+Quelle (grossteils) ((http://ultrahigh.org/2008/10/14/command-line-face-detection/))
  
-You will need the C++ Compiler 'g++' to make+You will need the C++ Compiler 'g++' to make. When not compiling enter: 
 +<code>apt-get install g++</code>
  
   * Download Torch3Vision and un-tar it: tar -zxf Torch3vision2.1.tgz   * Download Torch3Vision and un-tar it: tar -zxf Torch3vision2.1.tgz
Zeile 15: Zeile 19:
   * Build the vison examples for face detection: cd vision2.1/examples/facedetect/   * Build the vison examples for face detection: cd vision2.1/examples/facedetect/
  ../../../torch3make *.cc  ../../../torch3make *.cc
 +
 +
 +Jetzt 3 Programme in neuen Unterverzeichnis ''Linux_i686_dbg_float'':
 +  * haarscan -  Schnelle, zuverlässige Erkennung 
 +  * mlpcascadescan - Sehr zuverlössige Erkennung, bei ganz kleinen Bildern geht es nicht richtig, grosse (>3MPixel) dauert lange.
 +  * mlpscan - Seh keinen rechten Unterschied zum vorherigen.
 +
 +Jedes dieser 3 Programme braucht eine Modeldatei zur Gesichtererkennung. Diese 3 Dateien liegen unter ''vision2.1/examples/facedetect/models/''. Die jeweils passende Datei muss (mit Pfad) bei jeden Aufruf angegeben werden.
  
  
Zeile 25: Zeile 37:
 </code>  </code> 
 This command takes about 20s to run on my creaky old laptop, and creates two files. This command takes about 20s to run on my creaky old laptop, and creates two files.
 +
 +[[http://torch3vision.idiap.ch/examples2.0/|Source code of examples]]
 +
 +Parmaters (as far as i found out):
 +  * -minWsize n: min size n  in pixel to recognize a face
 +  * -nbest n : save the best n results only
 +  * -dir <string>                -> directory to store ouput files [.]
 +  * -savepos                     -> save pos file
 +  * -draw                        -> draw ppm image with detections
 +  * -savejpg                     -> save in jpeg instead of jpeg (jpeg ist wrong, should be ppm, but not working)
 +  * -verbose: tell a little bit more what you doing...
 +
 +The other file ‘andy.pos’ contains the results of face detection. Line one is the number of detections, then each line has format x y w h, very easy to parse.
 +<code>
 + FACE_POS=`head -n 2 “andy.pos | tail -n 1`
 + FACE_X=`echo $FACE_POS | awk ‘{print $1}’`
 + FACE_W=`echo $FACE_POS | awk ‘{print $3}’`
 + FACE_CENTER=`echo $FACE_X + $FACE_W/2 | bc` 
 +</code>
 +I played around with the step-factors in the x and y directions to shave a second or so off the face detection routine, the values I chose were 0.1 and 0.2 respectively (I don’t need any accuracy in the y direction really, since my use is to cut the face down the middle).
 +
 +Then, since these are portrait photographs, I can speed up face detection by setting a minimum size for the face. I experimented and one sixth of the total image width gave good results - any larger and the face detection would fail with a crash. Adding this constraint provides better than 10X speed up, since the algorithm doesn’t waste time searching for small faces.
 +<code>
 +WIDTH=`identify -format “%w” “andy.jpg”`
 + MIN_FACE_WIDTH=`echo $WIDTH / 6 | bc` 
 +</code>
 +So now here’s the final face detection command
 +<code>
 +mlpcascadescan “$ppm” -dir /tmp/ -savepos -model $MODEL \
 + -minWsize $MIN_FACE_WIDTH -stepxfactor $STEPX -stepyfactor $STEPY 
 +</code>
 +And finally, as promised, I’ll tell you how to blank-out one side of the face: of course, using Image Magick. Using the ‘chop’ or ‘crop’ commands didn’t work for this purpose, where I wanted the image to keep it’s dimensions but have one half just be white. So I decided to draw a white rectangle over half of the picture.  I apply the image manipulation to the original JPEG file, not the temporary PPM file that I used to detect the face position.
 +<code>
 +convert -fill white \
 + -draw “rectangle $FACE_CENTER,0 $WIDTH,$HEIGHT” \
 + “andy.jpg” “andy_half.jpg”
 +</code>
 +
 +Whole script:
 +
 +<code bash>
 +#!/bin/bash
 +
 +. ~/local/bin/die
 +
 +DRAW="-draw"
 +STEPX="0.05"
 +STEPY="0.25"
 +MODEL="$HOME/local/models/mlp-cascade19x19-20-2-110"
 +
 +# Find the center of the face in the image. Crop the image along the center of
 +# the face.
 +
 +# Requirements: 'torch3vision' http://torch3vision.idiap.ch/downloads.php
 +which mlpcascadescan > /dev/null || \
 + die "Please install mlpcascadescan from torch3vision"
 +
 +[ -e "$MODEL" ] || die "Model file $MODEL not found, get it from torch3vision"
 +
 +FACE_CENTER=""
 +
 +# Find the center of the face. Takes a jpeg and finds the face.
 +function find_center() {
 + jpeg="$1"
 + temp="/tmp/crop_center_face"
 + ppm="$temp.ppm"
 + pos="$temp.pos"
 + echo "Converting to ppm"
 + jpegtopnm $jpeg > $ppm
 + if [ -e "$pos" ]; then
 + rm "$pos"
 + fi
 +    HEIGHT=`identify -format "%h" "$jpeg"`;
 +    WIDTH=`identify -format "%w" "$jpeg"`;
 + MIN_FACE_WIDTH=`echo $WIDTH / 6 | bc`
 + echo "Beginning face detection, min-width=$MIN_FACE_WIDTH"
 + echo "mlpcascadescan "$ppm" -dir /tmp/ -savepos $DRAW -model $MODEL \
 + -minWsize $MIN_FACE_WIDTH -stepxfactor $STEPX -stepyfactor $STEPY"
 + mlpcascadescan "$ppm" -dir /tmp/ -savepos $DRAW -model $MODEL \
 + -minWsize $MIN_FACE_WIDTH -stepxfactor $STEPX -stepyfactor $STEPY
 + [ -e "$pos" ] || die "Face detection failed: $jpeg"
 + FACE_POS=`head -n 2 "$pos" | tail -n 1`
 + FACE_X=`echo $FACE_POS | awk '{print $1}'`
 + FACE_W=`echo $FACE_POS | awk '{print $3}'`
 + FACE_CENTER=`echo $FACE_X + $FACE_W/2 | bc`
 + echo "FACE_X: $FACE_X   FACE_W: $FACE_W   FACE_CENTER: $FACE_CENTER"
 +}
 +
 +function crop_image() {
 + jpeg="$1"
 + ext=${jpeg##*.}
 +    base=`basename "$jpeg" .$ext`
 +    crop="$base""_crop.jpg"
 +    HEIGHT=`identify -format "%h" "$jpeg"`;
 +    WIDTH=`identify -format "%w" "$jpeg"`;
 + echo "Cropping $jpeg at $FACE_CENTER, saving to $crop"
 + convert -fill white -draw "rectangle $FACE_CENTER,0 $WIDTH,$HEIGHT" "$jpeg" "$crop"
 +}
 +
 +
 +for file in $@; do
 + echo "Processing file: $file"
 + find_center $file
 + crop_image $file
 +done
 +</code>
 +
 +
 +
 +
 
Nach oben
face_detection/torch3vision.1243335324.txt.gz · Zuletzt geändert: (Externe Bearbeitung)
Übersicht Letzte Änderungen
chimeric.de = chi`s home Creative Commons License Valid CSS Driven by DokuWiki do yourself a favour and use a real browser - get firefox!! Recent changes RSS feed Valid XHTML 1.0
DFmW2CEce3htPL1uNQuHUVu4Tk6WXigFQp   Dogecoin Donations Accepted Here    DFmW2CEce3htPL1uNQuHUVu4Tk6WXigFQp  DFmW2CEce3htPL1uNQuHUVu4Tk6WXigFQp