Billiard-dataset

Billiard-dataset is a collection of recordings made for the scientific paper titled: "3D reconstruction system and multiobject local tracking algorithm designed for billiards". The main objective is the creation of a system to perform a reconstruction of billiards moves in a 3D virtual world and a new multi-object tracking algorithm applied to billiards.

How to cite this

More information about the aim of our work can be found in our paper. If you use this dataset, make sure you cite this work as: Latex:

@article{Rodriguez_Lozano_2023,
	doi = {10.1007/s10489-023-04542-3},
	url = {https://doi.org/10.1007%2Fs10489-023-04542-3},
	year = 2023,
	month = {jun},
	publisher = {Springer Science and Business Media {LLC}},
	author = {Francisco J. Rodriguez-Lozano and Juan C. G{\'{a}}mez-Granados and H{\'{e}}ctor Mart{\'{\i}}nez and Jose M. Palomares and Joaqu{\'{\i}}n Olivares},
	title = {3D reconstruction system and multiobject local tracking algorithm designed for billiards},
	journal = {Applied Intelligence}
}

IEEE: F. J. Rodriguez-Lozano, J. C. Gámez-Granados, H. Martínez, J. M. Palomares, and J. Olivares, “3D reconstruction system and multiobject local tracking algorithm designed for billiards,” Applied Intelligence, Jun. 2023, doi: 10.1007/s10489-023-04542-3.

APA: Rodriguez-Lozano, F. J., Gámez-Granados, J. C., Martínez, H., Palomares, J. M., & Olivares, J. (2023). 3D reconstruction system and multiobject local tracking algorithm designed for billiards. Applied Intelligence. https://doi.org/10.1007/s10489-023-04542-3

Dataset link

The dataset requires approximately 190 GB of disk space, and it is available at the following web Link.

Dataset content

The data set includes: 100 recordings of shots made in the blackball billiard modality; 100 recordings of shots in the carom billiard modality; 100 recordings of shots in the snooker billiard modality. In addition, not only are these recordings available, but the exact positions have been obtained manually for a subset of each modality, carrying out comparisons with different methodologies and with the algorithm proposed in our work. Therefore, in the dataset we can find the following folders and files:

• blackball:
  • blackball_[0...to...99]: subfolders of each recording of the blackball billiard modality.
    • BGR.dat: raw data of the recording. A special function is required to display it (see "How to read 'BGR.dat' raw data" section).
    • BGR.avi: rendered video from BGR.dat with compression.
    • AnimationsResults: subfolder with the 3D virtual generated worlds. It contains the animation files in ".x3d" format and rendered in video with format ".avi" with compression.
    • TrackingResults: subfolder with the results of all the tested algorithms. It contains "_track.png" images with the trajectories retrieved by each algorithm, and rendered videos comparing each algorithm.
  • background.png: a background image of the table without balls or external elements, in case any user needs to use it.
  • floderIndex.txt: A list of available subfolders.
• BlackBall_Manual:
  • BlackBall_Manual_[0...to...21]: subfolders of each recording of the blackball billiard modality.* This subfolder has inside the manual trajectory of the shot.
    • BGR.dat: raw data of the recording. A special function is required to display it (see "How to read 'BGR.dat' raw data" section).
    • BGR.avi: rendered video from BGR.dat with compression.
    • Jaccard_Coefficient.txt: a file that contains the results of the Jaccard index for each algorithm.
    • manual_track.png: the ground truth trajectories of each object in movement.
    • AnimationsResults: subfolder with the 3D virtual generated worlds. It contains the animation files in ".x3d" format and rendered in video with format ".avi" with compression.
    • TrackingResults: subfolder with the results of all the tested algorithms. It contains "track.png" images with the trajectories retrieved by each algorithm, "IOU.png" images with the Intersection over the Union of the trajectories of each algorithm with the Manual trajectory. Also, rendered videos comparing each algorithm are available.
  • background.png: a background image of the table without balls or external elements, in case any user needs to use it.
  • floderIndex.txt: A list of available subfolders.
  • Jaccard_Coefficient_Average.txt: an average of the results for each algorithm compared using the Jaccard index for the folders available in the file "folderIndex.txt".
• carom:
  • carom_[0...to...99]: subfolders of each recording of the carom billiard modality.
    • BGR.dat: raw data of the recording. A special function is required to display it (see "How to read 'BGR.dat' raw data" section).
    • BGR.avi: rendered video from BGR.dat with compression.
    • AnimationsResults: subfolder with the 3D virtual generated worlds. It contains the animation files in ".x3d" format and rendered in video with format ".avi" with compression.
    • TrackingResults: subfolder with the results of all the tested algorithms. It contains "_track.png" images with the trajectories retrieved by each algorithm, and rendered videos comparing each algorithm.
  • background.png: a background image of the table without balls or external elements, in case any user needs to use it.
  • floderIndex.txt: A list of available subfolders.
• Carom_Manual:
  • Carom_Manual_[0...to...15]: subfolders of each recording of the carom billiard modality.* This subfolder has inside the manual trajectory of the shot.
    • BGR.dat: raw data of the recording. A special function is required to display it (see "How to read 'BGR.dat' raw data" section).
    • BGR.avi: rendered video from BGR.dat with compression.
    • Jaccard_Coefficient.txt: a file that contains the results of the Jaccard index for each algorithm.
    • manual_track.png: the ground truth trajectories of each object in movement.
    • AnimationsResults: subfolder with the 3D virtual generated worlds. It contains the animation files in ".x3d" format and rendered in video with format ".avi" with compression.
    • TrackingResults: subfolder with the results of all the tested algorithms. It contains "track.png" images with the trajectories retrieved by each algorithm, "IOU.png" images with the Intersection over the Union of the trajectories of each algorithm with the Manual trajectory. Also, rendered videos comparing each algorithm are available.
  • background.png: a background image of the table without balls or external elements, in case any user needs to use it.
  • floderIndex.txt: A list of available subfolders.
  • Jaccard_Coefficient_Average.txt: an average of the results for each algorithm compared using the Jaccard index for the folders available in the file "folderIndex.txt".
• snooker:
  • snooker_[0...to...99]: subfolders of each recording of the snooker billiard modality.
    • BGR.dat: raw data of the recording. A special function is required to display it (see "How to read 'BGR.dat' raw data" section).
    • BGR.avi: rendered video from BGR.dat with compression.
    • AnimationsResults: subfolder with the 3D virtual generated worlds. It contains the animation files in ".x3d" format and rendered in video with format ".avi" with compression.
    • TrackingResults: subfolder with the results of all the tested algorithms. It contains "_track.png" images with the trajectories retrieved by each algorithm, and rendered videos comparing each algorithm.
  • background.png: a background image of the table without balls or external elements, in case any user needs to use it.
  • floderIndex.txt: A list of available subfolders.
• Snooker_Manual:
  • nooker_Manual_[0...to...15]: subfolders of each recording of the snooker billiard modality.* This subfolder has inside the manual trajectory of the shot.
    • BGR.dat: raw data of the recording. A special function is required to display it (see "How to read 'BGR.dat' raw data" section).
    • BGR.avi: rendered video from BGR.dat with compression.
    • Jaccard_Coefficient.txt: a file that contains the results of the Jaccard index for each algorithm.
    • manual_track.png: the ground truth trajectories of each object in movement.
    • AnimationsResults: subfolder with the 3D virtual generated worlds. It contains the animation files in ".x3d" format and rendered in video with format ".avi" with compression.
    • TrackingResults: subfolder with the results of all the tested algorithms. It contains "track.png" images with the trajectories retrieved by each algorithm, "IOU.png" images with the Intersection over the Union of the trajectories of each algorithm with the Manual trajectory. Also, rendered videos comparing each algorithm are available.
  • background.png: a background image of the table without balls or external elements, in case any user needs to use it.
  • floderIndex.txt: A list of available subfolders.
  • Jaccard_Coefficient_Average.txt: an average of the results for each algorithm compared using the Jaccard index for the folders available in the file "folderIndex.txt".

How to read 'BGR.dat' raw data

If the user wishes to use the original raw data (BGR.dat), he must adapt the following code to the specific needs of his program:

#include <fstream>
#include <iostream>
#include <opencv2/opencv.hpp>

using namespace std;
using namespace cv;

/*Function used to read the raw "BGR.dat" files using OpenCV and C++ */
int readMatFromFile(ifstream& input, Mat &Img)
{
int matCols = Img.cols, matRows = Img.rows;
int type = Img.type();
int cont=0, cont2=0;

Vec3b value;

  switch (type)
  {
    case CV_8UC3:
      for (cont=0; cont < matRows; ++cont)
      {
        for (cont2=0; cont2 < matCols; ++cont2)
        {  
          input.read((char*) &value, sizeof(value));
          Img.at<Vec3b>(cont, cont2) = value;           
        }
      }
    break;
    // more cases to add
    default:
      cout << "Error: format not supported! Conrrently supported: CV_8UC3\n";
      return -1;
    break;
  }

return 0;
}

int main (int argc, char* argv[])
{
ifstream BGRimage;
bool goOut = false;

  imagenRGB.open(argv[1], ios::in|ios::binary);

  while(!goOut)
  {
    Mat image(480, 640, CV_8UC3);
    // reading of each frame
    readMatFromFile(imagenRGB, image);

    // do something with the image
    
    if(imagenRGB.eof())
      goOut = true;
  }
  
  imagenRGB.close();  

return 0;
}

License

This dataset is licensed by the author under a CC BY-SA 4.0 license https://creativecommons.org/licenses/by-sa/4.0/deed.en.