This repository contains the binocular focal stack dataset in the paper: Deep eyes: Joint depth inference using monocular and
binocular cues. The dataset is generated based on FlyingThings3D.
binocular_focal_stack_clean/
train/ # Training scenes
[scene_id]_[focal_slice_id]_[L/R]_image.png # Image for a focal slice
test/ # Testing scenes (same layout as ./train/)
binocular_focal_stack_noise/
train/ # Training scenes
[scene_id]_[focal_slice_id]_[L/R]_image.png # Image for a focal slice, with Poisson noise added
test/ # Testing scenes (same layout as ./train/)
color_depth_pair/
train/ # Training scenes
[scene_id]_color_[L/R].png # All-focus color image
[scene_id]_depth_[L/R].png # GT disparity image
test/ # Testing scenes (same layout as ./train/)
For the disparity images, an intensity of 255 corresponds to a disparity of 100 pixels.
If you find our dataset or paper useful, please consider citing:
@article{chen2020deep,
title={Deep eyes: Joint depth inference using monocular and binocular cues},
author={Chen, Zhang and Guo, Xinqing and Li, Siyuan and Yang, Yang and Yu, Jingyi},
journal={Neurocomputing},
year={2020},
publisher={Elsevier}
}