   ***************************************************************
   **          Università  degli Studi di Parma (Italy)         **
   **       Dipartimento di Ingegneria dell'Informazione        **
   **   RIMLab - Robotics and Intelligent Machines Laboratory   **
   **              [http://rimlab.ce.unipr.it]                  **
   ***************************************************************
   ** caselli, aleotti, dlr, kallasi, oleari [@ce.unipr.it]     **
   ***************************************************************


______DESCRIPTION______________________________________________________________________

This underwater dataset has been taken near Portofino (Italy) on September 6th, 2014.
The dataset comprises image sequences of both single and multiple pipes with different colors and radii ranging from 5 to 6 cm. 
The depth was approximately 10 m and the dataset was acquired at 15 frames per second.
The dataset include 9600 stereo images in Bayer encoded format (1292x964 resolution), the original in-air camera calibration parameters, and the ones obtained after calibration with the method described in [oleari15system] (.yaml files).

The dataset is splitted in 3 separate archives in order to facilitate the download. 
Each archive is stand alone and contains a full set of images and calibration files.



______CONTENT__________________________________________________________________________

The dataset consists of the following directories:

-> CALIB: it contains two calibration sets of files for left and right cameras, for in-air calibration (ORIG_) and after the post-processing method (MOD_) described in [oleari15system].
          The YAML files are consistent with the OpenCV library format (see http://docs.opencv.org/modules/core/doc/xml_yaml_persistence.html)
    CALIB/ORIG_camera_info_left.yaml
    CALIB/ORIG_camera_info_right.yaml
    CALIB/MOD_camera_info_left.yaml
    CALIB/MOD_camera_info_right.yaml
-> IMGS: this directory contains pairs of images organized in three sets
    IMGS/set1
    IMGS/set2
    IMGS/set3



______REFERENCES_______________________________________________________________________

When using this dataset in your research, we will be happy if you cite us:

@inproceedings{oleari15system,
  author = {Oleari, F. and Kallasi, F. and Lodi Rizzini, D. and Aleotti, J. and Caselli, S.},
  title = {An Underwater Stereo Vision System: from Design to Deployment and Dataset Acquisition},
  booktitle = OCEANS,
  pages = {1-5},
  year = {2015},
}

@inproceedings{kallasi15graph,
  author = {Kallasi, F. and Lodi Rizzini, D. and Oleari, F. and Aleotti, J.},
  title = {Computer Vision in Underwater Environments: a Multiscale Graph Segmentation Approach},
  booktitle = OCEANS,
  pages = {1-6},
  year = {2015},
}

@article{lodirizzini15,
  author = {Lodi Rizzini, D. and Kallasi, F. and Oleari, F. and Caselli, S.},
  title = {Investigation of Vision-based Underwater Object Detection with Multiple Datasets},
  journal = IJARS,
  volume = {-},
  number = {-},
  pages = {-},
  month = {may},
  year = {2015},
  publisher = {InTech},
}



______ACKNOWLEDGMENT___________________________________________________________________

The system has been developed within the Italian national Project MARIS.
The authors would like to thank the divers of Federazione Italiana Attività Subacquee of Parma (FIAS, http://www.fiasparma.it/) for their valuable support in setting up the underwater image acquisition experiment.

