search

halajun / VDO_SLAM

VDO-SLAM: A Visual Dynamic Object-aware SLAM System
Other
729 stars 113 forks source link

how to get depth.png in kitti #43

Open andersonhusky opened 2 years ago

andersonhusky commented 2 years ago

hi! thanks for your excellent work , I have a quesion about how you get the depth.png under the /demo-kitti/depth?

MinaHenein commented 2 years ago

Hi @andersonhusky, This is explained in section IV. System in our paper VDO-SLAM: A Visual Dynamic Object-aware SLAM System (https://arxiv.org/pdf/2005.11052.pdf)

andersonhusky commented 2 years ago

thans for reply @MinaHenein! I know you are using MonoDepth2 to create depth image in Monocular mode, but it seems like the depth.png under the /demo-kitti/depth is calculated from Stereo mode

MinaHenein commented 2 years ago

@andersonhusky Yes, correct. From our paper:

The input to the system is stereo or RGB-D images. For stereo images, as a first step, we extract depth information by applying the stereo depth estimation method described in [62] to generate depth maps and the resulting data is treated as RGB-D.

[62] K. Yamaguchi, D. McAllester, and R. Urtasun, “Efficient Joint Segmentation, Occlusion Labeling, Stereo and Flow Estimation,” in European Conference on Computer Vision (ECCV). Springer, 2014, pp. 756–771.

andersonhusky commented 2 years ago

@MinaHenein got it! thanks again!