Skip to content

objects-in-space-and-time/v1like

BranchesTags
 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

54 Commits
v1like
v1like
 
 
.gitignore
.gitignore
 
 
README
README
 
 
requirements.txt
requirements.txt
 
 
setup.py
setup.py
 
 

Repository files navigation

Example of use:
===============

python v1like_extract_fromcsv.py -i ./test_imageset config/v1like_a.py ./test_imageset/train5test5_split01.csv .v1like_a  --nprocessors=$(cat /proc/cpuinfo | grep processor | wc -l)


Steps to reproduce the results from our PLoS 2008 paper:
========================================================

"Why is Real-World Visual Object Recognition Hard?"
By Nicolas Pinto, David D. Cox and James J. DiCarlo (2008)
Published in PLoS Comput Biol 4(1): e27.
doi:10.1371/journal.pcbi.0040027
http://www.ploscompbiol.org/article/info:doi/10.1371/journal.pcbi.0040027

Instructions:
-------------

# -- set up
export PJT=$HOME/plos08_reprod
export NPROCS=$(cat /proc/cpuinfo | grep processor | wc -l)
mkdir -p $PJT/{src,data}

cd $PJT/src
git clone https://github.com/npinto/v1like.git
git clone https://github.com/npinto/sclas.git
export V1LIKE=$PJT/src/v1like/v1like
export SCLAS=$PJT/src/sclas

# -- don't forget to install shogun
# see e.g.:
#
# * for Gentoo:
#   https://github.com/npinto/sekyfsr-gentoo-overlay
#   https://github.com/npinto/sekyfsr-gentoo-overlay/blob/master/sci-libs/shogun/shogun-0.9.3.ebuild
#
# * for Ubuntu 9.10:
#   https://github.com/npinto/np-toolbox/blob/master/install_scripts/install_shogun0.9.3_Ubuntu9.10.bash

# ----------------------------
# -- #Caltech101: get image set
cd $PJT/data
wget http://www.vision.caltech.edu/Image_Datasets/#Caltech101/101_ObjectCategories.tar.gz
tar xzvf 101_ObjectCategories.tar.gz

# -- #Caltech101: create splits
export ntrain=15 ntest=15
for i in `seq -w 1 10`; do
    python $SCLAS/create_traintest_split.py --rseed=$i --ntrain=$ntrain --ntest=$ntrain $PJT/data/101_ObjectCategories/{,train${ntrain}test${ntest}_split_${i}.csv};
done;

# -- #Caltech101: generate v1like features
for conf in v1like_a{,_plus}; do
    for i in `seq -w 1 10`; do
        python $V1LIKE/v1like_extract_fromcsv.py --nprocessors=$NPROCS -i $PJT/data/101_ObjectCategories/ $V1LIKE/config/$conf.py $PJT/data/101_ObjectCategories/train${ntrain}test${ntest}_split_${i}.csv $conf.mat;
    done;
done;

# -- #Caltech101: generate kernels
for conf in v1like_a{,_plus}; do
    for csv in $PJT/data/101_ObjectCategories/train${ntrain}test${ntest}_split_??.csv; do
        python $SCLAS/kernel_generate_fromcsv.py -i $(dirname $csv) $csv $conf.mat $csv.kernel.$conf.mat;
    done;
done;

# -- #Caltech101: run SVMs
for conf in v1like_a{,_plus}; do
    for csv in $PJT/data/101_ObjectCategories/train${ntrain}test${ntest}_split_??.csv; do
        python $SCLAS/svm_ova_fromfilenames.py $csv.kernel.$conf.mat -o $csv.svm_ova_results.$conf.mat;
    done;
done;

# -- #Caltech101: average classification results (crudely ;-)
for conf in v1like_a{,_plus}; do
    echo $conf;
    for i in `seq -w 1 10`; do
        python $SCLAS/print_mat.py $PJT/data/101_ObjectCategories/train${ntrain}test${ntest}_split_${i}.csv.svm_ova_results.$conf.mat accuracy;
    done | awk '{sum+=$2} END {print sum/NR}';
done;
# v1like_a
# 58.0392
# v1like_a_plus
# 61.4183

# ----------------------------------------
# -- Controlled Invariance: get image sets
cd $PJT/data
wget http://s3.amazonaws.com/PLoS08_ControlSets/PLoS08_ControlSet_Cars_Planes_v01.tar.gz http://s3.amazonaws.com/PLoS08_ControlSets/PLoS08_ControlSet_Cars_Planes_v01.tar.gz.md5
md5sum -c PLoS08_ControlSet_Cars_Planes_v01.tar.gz.md5
tar xzvf PLoS08_ControlSet_Cars_Planes_v01.tar.gz

# -- Controlled Invariance: create splits
export ntrain=100 ntest=30
for dir in $(ls -d $PJT/data/PLoS08_ControlSet_Cars_Planes_v01/*/); do
    for i in `seq -w 1 10`; do
        python $SCLAS/create_traintest_split.py --rseed=$i --ntrain=$ntrain --ntest=$ntest $dir/{,train${ntrain}test${ntest}_split_${i}.csv};
    done;
done;

# -- Controlled Invariance: generate v1like features
for conf in v1like_a{,_plus}; do
    for dir in $(ls -d $PJT/data/PLoS08_ControlSet_Cars_Planes_v01/*/); do
        for i in `seq -w 1 10`; do
            python $V1LIKE/v1like_extract_fromcsv.py --nprocessors=$NPROCS -i $dir/ $V1LIKE/config/$conf.py $dir/train${ntrain}test${ntest}_split_${i}.csv $conf.mat;
        done;
    done;
done;

# -- Controlled Invariance: generate kernels
for conf in v1like_a{,_plus}; do
    for dir in $(ls -d $PJT/data/PLoS08_ControlSet_Cars_Planes_v01/*/); do
        for csv in $dir/train${ntrain}test${ntest}_split_??.csv; do
            python $SCLAS/kernel_generate_fromcsv.py -i $(dirname $csv) $csv $conf.mat $csv.kernel.$conf.mat;
        done;
    done;
done;

# -- Controlled Invariance: run SVMs
for conf in v1like_a{,_plus}; do
    for dir in $(ls -d $PJT/data/PLoS08_ControlSet_Cars_Planes_v01/*/); do
        for csv in $dir/train${ntrain}test${ntest}_split_??.csv; do
            python $SCLAS/svm_ova_fromfilenames.py $csv.kernel.$conf.mat -o $csv.svm_ova_results.$conf.mat;
        done;
    done;
done;

# -- Controlled Invariance: average classification results (crudely ;-)
for conf in v1like_a{,_plus}; do
    for dir in $(ls -d $PJT/data/PLoS08_ControlSet_Cars_Planes_v01/*/); do
        echo $conf $dir;
        for i in `seq -w 1 10`; do
            python $SCLAS/print_mat.py $dir/train${ntrain}test${ntest}_split_${i}.csv.svm_ova_results.$conf.mat accuracy;
        done | awk '{sum+=$2} END {print sum/NR}';
    done;
done;
#v1like_a plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation00_NaturalBg_n130_200x200_GS/
#97.1667
#v1like_a plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation00_NoiseBg_n130_200x200_GS/
#100
#v1like_a plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation00_PhaseScrambledBg_n130_200x200_GS/
#99.8333
#v1like_a plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation01_NaturalBg_n130_200x200_GS/
#87.3333
#v1like_a plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation01_NoiseBg_n130_200x200_GS/
#97.8333
#v1like_a plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation01_PhaseScrambledBg_n130_200x200_GS/
#92.5
#v1like_a plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation02_NaturalBg_n130_200x200_GS/
#68.3333
#v1like_a plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation02_NoiseBg_n130_200x200_GS/
#84
#v1like_a plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation02_PhaseScrambledBg_n130_200x200_GS/
#81.3333
#v1like_a plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation03_NaturalBg_n130_200x200_GS/
#65.5
#v1like_a plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation03_NoiseBg_n130_200x200_GS/
#74.1667
#v1like_a plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation03_PhaseScrambledBg_n130_200x200_GS/
#65.6667
#v1like_a plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation04_NaturalBg_n130_200x200_GS/
#54.8333
#v1like_a plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation04_NoiseBg_n130_200x200_GS/
#68
#v1like_a plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation04_PhaseScrambledBg_n130_200x200_GS/
#62.5
#v1like_a plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation05_NaturalBg_n130_200x200_GS/
#51.3333
#v1like_a plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation05_NoiseBg_n130_200x200_GS/
#61.8333
#v1like_a plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation05_PhaseScrambledBg_n130_200x200_GS/
#57.5
#v1like_a plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation06_NaturalBg_n130_200x200_GS/
#45.5
#v1like_a plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation06_NoiseBg_n130_200x200_GS/
#55
#v1like_a plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation06_PhaseScrambledBg_n130_200x200_GS/
#50.6667
#v1like_a_plus plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation00_NaturalBg_n130_200x200_GS/
#97.3333
#v1like_a_plus plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation00_NoiseBg_n130_200x200_GS/
#100
#v1like_a_plus plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation00_PhaseScrambledBg_n130_200x200_GS/
#99.6667
#v1like_a_plus plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation01_NaturalBg_n130_200x200_GS/
#86
#v1like_a_plus plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation01_NoiseBg_n130_200x200_GS/
#98.3333
#v1like_a_plus plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation01_PhaseScrambledBg_n130_200x200_GS/
#92.3333
#v1like_a_plus plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation02_NaturalBg_n130_200x200_GS/
#67.1667
#v1like_a_plus plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation02_NoiseBg_n130_200x200_GS/
#88.6667
#v1like_a_plus plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation02_PhaseScrambledBg_n130_200x200_GS/
#80.6667
#v1like_a_plus plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation03_NaturalBg_n130_200x200_GS/
#63.8333
#v1like_a_plus plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation03_NoiseBg_n130_200x200_GS/
#78.5
#v1like_a_plus plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation03_PhaseScrambledBg_n130_200x200_GS/
#69.1667
#v1like_a_plus plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation04_NaturalBg_n130_200x200_GS/
#51.5
#v1like_a_plus plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation04_NoiseBg_n130_200x200_GS/
#70.6667
#v1like_a_plus plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation04_PhaseScrambledBg_n130_200x200_GS/
#64.5
#v1like_a_plus plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation05_NaturalBg_n130_200x200_GS/
#51.1667
#v1like_a_plus plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation05_NoiseBg_n130_200x200_GS/
#64.8333
#v1like_a_plus plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation05_PhaseScrambledBg_n130_200x200_GS/
#58.1667
#v1like_a_plus plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation06_NaturalBg_n130_200x200_GS/
#47
#v1like_a_plus plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation06_NoiseBg_n130_200x200_GS/
#57.5
#v1like_a_plus plos08_reprod/data/PLoS08_ControlSet_Cars_Planes_v01/ControlSet_Cars_Planes_Variation06_PhaseScrambledBg_n130_200x200_GS/
#55.3333

About

web.mit.edu/pinto

Resources

Readme
Activity
Custom properties

Stars

1 star

Watchers

2 watching

Forks

0 forks
Report repository

Releases

1 tags

Packages

No packages published

Languages

  • Python 100.0%