OpenBot AMSR Files

JPR - July 17, 2011 - Vers. 0.35
New features added to: facerecognition-gtk:
---------------------------------------
1 - Face identification numbers are displayed next to the faces detected.
2 - faces id are recorded in the file 'facedetect.data'
File structure: 
timestamp-duration nb_faces-largest_face# x_coordinate:y_coordinate widthxheight horizontal_move:front_back_move id_face#1 id_face#2 ...
Example:
1310824063-1 2-3 324:217 146x146 0:0 1310740601 1310751234

Note:   timestamp is the number of seconds since 1970/01/01
        id_face#1,id_face#2,... are numbers attributed to a new face detected, 
        there're in fact timestamp when the shoot button was clicked.
3 - Option --autoshoot=1 allows autmomatic pictures shooting and Eigenfaces training when an unknown face is detected.
4 - A tiny label (next the Shoot button) gives the learning status: 
    'SHOOT' (30 frames are recorded)  'TRAIN' (the training prog runs in background)  '.....' (no learning)

JPR - July 15, 2011 - Vers. 0.3
New program: facerecognition-gtk:
---------------------------------------
facerecognition-gtk is facedetector-gtk with a face recognition added, so facerecognition-gtk 
provides exactly the same functionalities as facedetector-gtk plus the recognition. 
facerecognition-gtk is experimental and intended for robotic applications.

1 - recognize face (use eigenface for face recognition)

2 - take 30 pictures (shoot button) and automatically train the classifier with 
facerecognition-train (the program runs in backgound during the training).

How to run it:
- Start facedetector-gtk from the program directory, for instance:
  ./facerecognition-gtk --algorithm=2 --sizemin=120 --posmax=80 --graphical=1 --debug=1
- When your face is detected (red rectangle around it), click on shoot and do not move too much for 3 seconds.
Then the program will take 30 frames that will be saved in faces/new/ directory and will launch facedetector-train.
The eigenface face classifier is automatically trained and the face will have an identification number attributed.
The training takes a few seconds and as soon as the training done, facedetectector-gtk will recognize the new face 
( the id number will be displayed in the bottom of the image ).
the directory faces/s0 contains the faces trained.
the file train.txt contains the id and face filenames


JPR - June 30, 2011 - Vers. 0.2
New features added to facedetector-gtk:
---------------------------------------
1 - Detect the main face moving forward/backward (additionaly to moving left and right).

2 - Reject too small faces and reject coordinates moving too quickly can be modified (spin buttons).

3 - Record data of the face detected (timestamp,nb of faces,coordinates,sizes,direction) in a file,
 (by default the filename is facedetect.data and is located in the current directory), 
File structure: timestamp-duration nb_faces-face# x_coordinate:y_coordinate widthxheight horizontal_move:front_back_move

4 - Text results concerning the biggest face detected are recorded (timestamp,nb of faces,coordinates,sizes,direction) in a file,
 (by default the filename is facedetect.data and is located in the current directory), 
the result recorded is continuouly updated and is designed to be used by "classify" (the decision prog)
File structure: nb of faces detected - moving direction
Example of detection recorded:
no face detected
1 face(s) detected - right front moving
1 face(s) detected - no  moving
1 face(s) detected - left back moving


5 - Command line options available:
    --image="<path>/file_name"      //to analyze a picture
    --file1="<path>/file_name"       //to record result in another directory/filename
    --file2="<path>/file_name"       //to record result in another directory/filename
    --algorithm=1..99               //algorithm 1..4 are face detection
    --sizemin=1..200                //minimum size of the face (in pixels)
    --posmax=1..200                 //maximun moving allowed (in pixels)
    --graphical=0 or 1              //0 to display only the images and faces detected
    --debug=0 or 1                  //1 to append each result
    --help or -h 

JPR - June 27, 2011 - Vers. 0.15
facedetector-gtk : modified, new features added
- Reject too small faces (< 50 pixels,hard coded lines 188 and 189,limit.width = 50;limit.height = 50;)
- Reject coordinates moving too quickly between 2 pictures (hard coded line 190, limit.pos = 50;)
- Detect the moving direction (left / right)

JPR - June 25, 2011
Some change to compile on Ubuntu 11.04 x64.
facedetector-gtk : several bugs fix and code cleanup