search

fossasia / visdom

A flexible tool for creating, organizing, and sharing visualizations of live, rich data. Supports Torch and Numpy.
Apache License 2.0
9.99k stars 1.14k forks source link

Large scale 3D point cloud visualization pane #686

Open AkiraHero opened 4 years ago

AkiraHero commented 4 years ago

Is your feature request related to a problem? Please describe. It will be convenient if Visdom support point cloud visualization in deep learning about 3D vision, like project using PointNet.

Describe the solution you'd like Considering large scale point cloud collected for autonomous driving or 3D reconstruction, a visualization using WebGL maybe a good choice, which supports mouse interaction.

Describe alternatives you've considered A clear and concise description of any alternative solutions or features you've considered.

Additional context Add any other context or screenshots about the feature request here.

JackUrb commented 4 years ago

We don't currently have any plans to extend beyond the support of what Plotly's 3d scatterplots (which do have webgl support) are able to manage. Let us know if the webgl-backed plotly plots are enough to support your case.

If not, if you're interested in scoping out options for supporting even larger scale 3d point plots please put any implementation ideas you may find here (existing libraries, etc.) so that anyone interested in contributing to this issue would be able to work on it.

mjstevens777 commented 4 years ago

@AkiraHero I have made a wrapper around THREE.js to display point clouds and other 3D shapes in visdom (as well as jupyter notebooks). It takes numpy arrays as input, converts them into HTML/javascript, and renders the result using viz.text(). Let me know if you would like me to share.

I imagine it would also be pretty easy to extend something like pythreejs to do the same thing. (See ipywidgets)