This code is based on the paper A Robust Indoor Scene Recognition Method based on Sparse Representation presented on the 22nd Iberoamerican Congress on Pattern Recognition (CIARP 2017). The goal of this software is to build a robust representation of scene images, that conveys global as well as local information, for the task of scene recognition. We built an over-complete dictionary whose base vectors are feature vectors extracted from fragments of a scene, and the final representation of an image is a linear combination of the visual features of objects’ fragments.

For more information, please access the project page.

• Guilherme Nascimento - MSc student - UFMG - guigonasc@gmail.com
• Camila Laranjeira - MSc student - UFMG - mila.laranjeira@gmail.com
• Vinicius Braz - Undergraduate Student - UFMG - viniciusbraz30@gmail.com
• Anisio Lacerda - Co-Avisor - CEFET-MG - anisiom@gmail.com
• Erickson R. Nascimento - Advisor - UFMG - erickson@dcc.ufmg.br

Federal University of Minas Gerais (UFMG)
Computer Science Department
Belo Horizonte - Minas Gerais -Brazil

VeRLab: Laboratory of Computer Vison and Robotics
http://www.verlab.dcc.ufmg.br

• OpenCV 3.2.0
• Caffe 1.0.0
• Spams 2.5
• CUDA 8.0 (GPU version only)
• cuDNN v5 (GPU version only)
• Numpy, Scikit Image, Scikit Learn, Scipy, Pickle

1. Generate Train/Test split:
   Example provided in folder CFG_FILES (fold4.cfg). The source code also contains the script kfold.py if you wish to generate new splits.

2. Edit Config files:
   Examples provided in folder CFG_FILES. Config files should be stored in the same path as the on provided in --folder parameter (as seen later). The code requires two files:

• IMNET.cfg: referring to VGG16 trained on ImageNet
• PLACES.cfg: referring to VGG16 trained on Places205

3. Execution:
   Execute run_test.py using the following parameters:

• -f, --folder: Path to folder you wish to save the outputs of the code;
• -o, --output: Path to file you wish to save the output statistics (e.g. accuracy);
• -k, --fold: Index of Train/Test split (referring to the parameter [folds] in the Config files);
• -m, --mode: Operation mode ('train' or 'test');
• -d, --ns1: Size of dictionary for scale 1;
• -e, --ns2: Size of dictionary for scale 2;
• -l, --lambda: Sparsity (e.g. 0.1 to activate at most 10% of the dictionary);
• -t, --method: Minimization Method ('OMP', 'SOMP' or 'LASSO');
• -j, --dl: Sparsity controller for dictionary learning.

Example of Usage:

python run_test.py -f /root/output -o /root/output/result_ -k 4 -m train -d 603 -e 3283 -l 0.1 -t OMP -j 0.03 

If you are using it for academic purposes, please cite:

G. Nascimento, C. Laranjeira, V. Braz, A. Lacerda, E. R. Nascimento, A Robust Indoor Scene Recognition Method based on Sparse Representation, in: 22nd Iberoamerican Congress on Pattern Recognition, CIARP, Springer International Publishing, Valparaiso, CL, 2017. To appear.

@inproceedings{Nascimento2017,
Title = {A Robust Indoor Scene Recognition Method based on Sparse Representation},
Author = {Nascimento, Guilherme and Laranjeira, Camila and Braz, Vinicius and Lacerda, Anisio and Nascimento, Erickson Rangel},
booktitle = {22nd Iberoamerican Congress on Pattern Recognition. CIARP},
Publisher = {Springer International Publishing},
Year = {2017},
Address = {Valparaiso, CL},
note = {To appear},
}