Skip to content


Subversion checkout URL

You can clone with
Download ZIP
Code release for sparselet release 1
Matlab C++ Python M
Branch: master

hos dirs

latest commit bf7977c2b8
rksltnl authored
Failed to load latest commit information.
VOC2007 hos: initial commit
bin hos dirs
data hos: initial commit
features hos: initial commit
gdetect hos dirs
model hos: initial commit
sparselets hos dirs
test hos dirs
utils hos: initial commit
vis hos: initial commit
000313.jpg hos: initial commit
000358.jpg hos: initial commit
001998.jpg hos: initial commit
004637.jpg hos: initial commit
LICENSE hos hos
compile.m hos
convolution_data.mat hos: initial commit
demo_detection.m hos: initial commit
demo_saved_feb18.m hos: initial commit
startup.m hos
voc_config.m hos: initial commit

README for sparselet code


Citing sparselets

If you find sparselet detection code useful in your research, please cite:

    title = "Generalized Sparselet Models for Real-Time Multiclass Object Recognition",
    booktitle = "IEEE Transactions on Pattern Analysis and Machine Intelligence",
    year = "2014",
    author = "Hyun Oh Song and Ross Girshick and Stefan Zickler and Christopher Geyer and Pedro Felzenszwalb and Trevor Darrell",

    title = "Discriminatively Activated Sparselets",
    booktitle = "International Conference on Machine Learning (ICML)",
    year = "2013", 
    author = "Ross Girshick and Hyun Oh Song and Trevor Darrell",

@inproceedings {Song-ECCV2012,
    title = "Sparselet Models for Efficient Multiclass Object Detection",
    booktitle = "European Conference on Computer Vision (ECCV)",
    year = "2012",
    author = "Hyun Oh Song and Stefan Zickler and Tim Althoff and Ross Girshick and Mario Fritz and Christopher Geyer and Pedro Felzenszwalb and Trevor Darrell",


Sparselet is released under the Simplified BSD License (refer to the LICENSE file for details).

System Requirements

Install instructions

  1. Download and install Intel® C++ Composer XE from the link above.
  2. Unpack the sparselet code.
  3. Download and install SPAMS toolbox in the same directory level as in the sparselet code.
  4. On a terminal run $python sparselets/ (for OS X) or $python sparselets/ (for Linux)
  5. Start matlab.
  6. Run the 'compile' function to compile the helper functions. (you may need to edit compile.m to use a different convolution routine depending on your system)
  7. Use 'demo_detection' code for a demo usage of the sparselet code for multiclass object detection.
Something went wrong with that request. Please try again.