OBJECT BANK RELEASE 1.0 (beta)

The code is tested on gcc version 4.1.2 and Linux distribution version CentOS release 5.5 (Final)

TO BUILD: 
type 'make clean; make'

REQUIREMENTS:
Libjpeg for loading images

QUICK TEST:
./OBMain ./inputs/ ./outputs/
This will run a test on a sample image from UIUC Sports dataset.

INSTRUCTIONS FOR RUNNING DIRECTLY:

./OBMain [-M file] [-R 0/1] [-F 0/1] [-L:#] inputdirectory/ outputdirectory/

-M: If you pass in a file that contains the model names, the code will only run that subset
-R: Indicate if you want to save the responsemaps (default: NO)
-F: Indiate if you want to save the features (default: YES)
-L: Indicate how many scale levels you want to descend into the image (default: 6)
Note: Be sure to leave the trailing '/' for the directories.

INSTRUCTIONS FOR RUNNING WITH SCRIPTS:
The two scripts included are RunDir.sh and ClusterExample.sh

The usage here for RunDir is:
./Rundir.sh [input directory that contains multiple folders of images] [base directory for outputs]
Example:
./RunDir.sh ../UIUC_SPORTS/ outputs/

So this would sequentially call the following
./OBmain ../UIUC_SPORTS/badminton/ outputs/badminton/
./OBmain ../UIUC_SPORTS/rowing/ outputs/rowing/
./OBmain ../UIUC_SPORTS/badminton/croquet outputs/croquet/
...etc.

For parallelization, we included ClusterExample.sh. The usage is:
 ./ClusterExample.sh dirlist ../UIUC_Sports/ outputs/ 1

The first input is a file containing all the folder names inside the input directory 
	(in the above example ../Indoor_images/)
The second input in the base directory for the inputs
The third inputs in the base directory for the outputs
The fourth input is the index for exactly which entry in of the dirlist you want to access.
	NOTE: This is a 0-indexed script! 

For example, the above example would call:
./OBmain ../UIUC_SPORTS/badminton/ outputs/badminton/

POST PROCESSING:
If you have a directory of categorized outputs that you would like to format for testing, we provide 
two MATLAB scripts for your convenience. Please refer to CreateFeatureVector.m 
For example

CREATING .TEXT MODEL FILES:
If you have additional LSVM models that you would like to include, you need can convert them to ASCII 
using the ModelMat2Txt.m. Copy the resulting text model into the models/ directory and it will be 
included in the next execution of OBmain

feature extraction using this version of code takes around 10 seconds per image.
