ICML 2013 experiments for "Consistency of Online Random Forests" using rftk
Switch branches/tags
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
experiment_config Icml 2013: compare to offline experiment working with refactored code May 15, 2013
experiment_utils
figures
results
source_data
.gitignore
README.md
dist_utils.py
inspect_result_pickle.py
kinect_bundle_data.py
kinect_measure_accuracy.py
kinect_plot_results.py
kinect_train_offline.py
kinect_train_online.py
kinect_train_online_single_images.py
kinect_train_saffari.py
kinect_utils.py
plot_advantage_of_forest.py
plot_compare_to_offline.py
plot_result_tensor.py
plot_utils.py
run_advantage_of_forest.sh
run_compare_to_offline.sh
run_kinect.sh
train_offline.py
train_online.py
train_saffari.py

README.md

rftk-colrf-icml2013

ICML 2013 experiments for "Consistency of Online Random Forests" using rftk. To run the experiments you must clone and build rftk and rftk-colrf-icml2013. The rftk project has undergone significant refactoring since the paper was published. We recommend using the latest implementation but we also provide the commits required to reproduce the exact results from the paper.

Setup

Clone both projects

git clone https://github.com/david-matheson/rftk.git
git clone https://github.com/david-matheson/rftk-colrf-icml2013.git

If you wish to use the exact setup as the paper checkout the colrf-icml2013-camera-ready tag.

cd /path/to/rftk
git checkout -b colrf-icml2013-camera-ready origin/colrf-icml2013-camera-ready
cd /path/to/rftk-colrf-icml2013
git checkout -b colrf-icml2013-camera-ready origin/colrf-icml2013-camera-ready

However, we recommend using the latest implementation in the colrf-icml2013 branch.

cd /path/to/rftk
git checkout -b colrf-icml2013 origin/colrf-icml2013
cd /path/to/rftk-colrf-icml2013
git checkout -b colrf-icml2013 origin/colrf-icml2013

Build rftk

cd /path/to/rftk
scons

Add rftk to your PYTHONPATH

PYTHONPATH=/path/to/rftk/
export PYTHONPATH

Download and unzip data

cd /path/to/rftk-colrf-icml2013/source_data
wget http://www.cs.ubc.ca/~mdenil/projects/random_forests/usps_mog.tar.bz2
tar -jxvf usps_mog.tar.bz2
wget http://www.cs.ubc.ca/~mdenil/projects/random_forests/kinect_data.zip
unzip kinect_data.zip

Running experiments

cd /path/to/rftk-colrf-icml2013/
./run_advantage_of_forest.sh
./run_compare_to_offline.sh
./run_kinect.sh