We share the mDFF dataset and its toolbox in this repository. Dataset is preprocessed as follows:
- we align and scale/crop the images using the alignment script.
- we project the pointclouds to the 2D image plane given the camera parameters using the projection script.
- we generate the train/val and test datasets in .h5 file using the dataset generation script.
Cropped (224x224) and clipped (0.1 to 4meters) hdf5 dataset can be downloaded here. It contains the train, val and test data.
Original focalstacks (/focalstacks), preprocessed focalstacks (/test, /train), registered depth maps and raw pointclouds can be downloaded here.
All data in the mDFF dataset is licensed under a Creative Commons 4.0 Attribution License (CC BY 4.0).
Caner Hazirbas, Sebastian Georg Soyer, Maximilian Christian Staab, Laura Leal-Taixé and Daniel Cremers, "Deep Depth From Focus", ACCV, 2018. (arXiv)
@InProceedings{hazirbas18ddff,
author = {C. Hazirbas and S. G. Soyer and M. C. Staab and L. Leal-Taixé and D. Cremers},
title = {Deep Depth From Focus},
booktitle = {Asian Conference on Computer Vision (ACCV)},
year = {2018},
month = {December},
eprint = {1704.01085},
url = {https://hazirbas.com/projects/ddff/},
}
The toolbox is released under GNU General Public License Version 3 (GPLv3).