Closed aminzabardast closed 5 years ago
The README.md
file of dataset clears up the GT issue.
The EndoAbS Dataset (Endoscopic Abdominal Stereo Images Dataset) aims to provide the computer assisted surgery community with a dataset for the validation of 3D reconstruction algorithms.
It is composed of:
The images were captured under different conditions:
If you use this dataset, please cite:
A.S. Ciullo, V. Penza, L. Mattos, E. De Momi
"Development of a surgical stereo endoscopic image dataset for
validating 3D stereo reconstruction algorithms." 6th Joint Workshop on
New Technologies for Computer/Robot Assisted Surgery. 2016.
We will soon ulpoad the more detailed Journal paper!
The EndoAbS dataset is organized as described in the following lines:
|_EndoAbS_dataset/ |_CalibrationParameters1.xml (*) |CalibrationParameters_2.xml (**) |__README |ORGAN_NAME/ |_distance_max/ |_distancemin/ |pose1/ |_imgL_1.png (low brightness) |_imgR1.png |imgL_2.png (medium brightness) |_imgR2.png |imgL_3.png (high brightness) |_imgR3.png |imgL_4.png (smoke) |_imgR4.png |imgL_5.png (smoke) |_imgR5.png |imgL_6.png (smoke) |_imgR6.png |point_map.xml (GT information) ... |___posen/
** The file 'CalibrationParameters_2.xml' contains intrinsic and extrinsic camera calibration parameters for all the others images (distance_min for liver and distance_max for kidney, liver and spleen).
The calibration files use the OpenCV XML format with the following data:
'Ml' and 'Mr' are the intrinsic parameters, respectively for left and right camera, in this format:
fx 0 cx
0 fy cy
0 0 1
'Dl' and 'Dr' are the distortion parameters, respectively for left and right camera.
'R' and 'T' are the extrinsic parameters, respectively rotation matrix and translation matrix.
GT files contain (x,y,z) triplets per line, stored in column major order. (0,0,0) triplet means that no GT was found. From this list, the corresponding GT 2D map can be calcuated, where each cell corresponds to an image pixel.
For further information, do not hesitate to contact veronica.penza@polimi.it.
Try and convert the Point Cloud
to Disparity
.
Note: Authors say it is possible.