Fast and Robust Multiple ColorChecker Detection

Paper | arXiv

Pedro D. Marrero Fernandez¹, Fidel A. Guerrero-Peña¹, Tsang Ing Ren¹, Jorge J. G. Leandro²

¹Universidade Federal de Pernambuco, ²Motorola Mobility LLC, a Lenovo Company

In Image and Vision Computing 2018

This code implements a multiple colorChecker detection method, as described in the paper Fast and Robust Multiple ColorChecker Detection. The process is divided into two steps: (1) ColorCheckers localization and (2) ColorChecker patches recognition.

Pipeline

Results of MCCNetFind for the synthetic dataset

Results of MCCNetFind for the real dataset

Requirements

You need OpenCV v3.1.0 or later and NVidia Caffe. This installation package contains support for opencv compilation for Windows in vs.12, vs.14 and mingw.

apt-get update
apt-get upgrade
nvcaffe
echo "export PYTHONPATH=/opt/nvcaffe/python" >> ~/.bashrc
exho "export PATH=/opt/nvcaffe/build/tools:/opt/nvcaffe/python:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin " >> ~/.bashrc
source ~/.bashrc
opencv

Installation

Building the project using CMake from the command-line:

export OpenCV_DIR="./extern/opencv"
mkdir build
cd build
cmake -D OpenCV_DIR=$OpenCV_DIR ..
make

How to use it

Preparing the chart

You should have done a photo or video of the ColorChecker Passport.

Photo of the ColorChecker Passport

Running the MCCFind:

./build/src/mcc ../db/img-colorchecker.jpg -o=../out -t=1 -sh -gt -nc=0
./build/src/mcc ../db/vdo-colorchecker.mp4 -o=../out -t=2 -sh -gt -nc=2
./build/src/mcc ../db/sec-colchecker-0.jpg -o=../out -t=3 -sh -gt -nc=2 -me=10.0

options:

 -t   # application type - 1 single image, 2 video, 3 image sequence
 -o   # output dir - default current dir
 -me  # minimum error
 -nc  # number maximum of checker color in the image
 -sh  # show result
 -gt  # generate table .csv format
  []  # input dir

Running the MCCFindNet:

usage: mccfindnet.py [-h] --configurate C [--no-gpu] [--json] [--no-show]
                 [--draw-cam] [--camdevice N] [--border N]

Citation

If you find this useful for your research, please cite the following paper.

@article{MARREROFERNANDEZ2018,
   title = "Fast and Robust Multiple ColorChecker Detection using Deep Convolutional Neural Networks",
   journal = "Image and Vision Computing",
   year = "2018",
   issn = "0262-8856",
   doi = "https://doi.org/10.1016/j.imavis.2018.11.001",
   url = "http://www.sciencedirect.com/science/article/pii/S0262885618301793",
   author = "Pedro D. Marrero Fernández and Fidel A. Guerrero Peña 
             and Tsang Ing Ren and Jorge J.G. Leandro",
}

Acknowledgments

This work was supported by the research cooperation project between Motorola Mobility (a Lenovo Company) and CIn-UFPE. Tsang Ing Ren, Pedro D. Marrero Fernandez and Fidel A. Guerrero-Peña gratefully acknowledge financial support from the Brazilian government agency FACEPE. The authors would also like to thank Leonardo Coutinho de Mendonça, Alexandre Cabral Mota, Rudi Minghim and Gabriel Humpire for valuable discussions.

pedrodiamel / colorchecker-detection

readme