Closed Yiiii19 closed 11 months ago
Hi @Yiiii19. We haven't got a tutorial on this. For custom RGBD dataset, we need aligned depth and color images, usually depth is aligned to color, together with poses. If ground truth poses are not available, the pipeline can anchor the world coordinate based on first frame, in this case, the value should be changed to 0 here as currently first two poses are initialized using ground truth.
There's a tutorial on capturing RealSense data with Open3D. I can't provide more help on this. Once dataset is ready, the left effort would be on new dataset class to load depth, color and pose as in line here.
Hi, thanks for the amazing work! I would like to ask if you have any tutorial for running the reconstruction on custom dataset? I have Realsense Camera and running open3D in advance is not a problem as preprocess. Thanks!