hankyang94 / damp

DAMP: Dynamical Pose Estimation

Dynamical Pose Estimation

This repository is the official Matlab implementation of Dynamical Pose Estimation, which has been accepted to be published in the International Conference on Computer Vision (ICCV), 2021.

If you find this implementation useful for your research, please cite:

@InProceedings{Yang21ICCV-DAMP,
    title={Dynamical Pose Estimation},
    author={Yang, Heng and Doran, Chris and Slotine, Jean-Jacques},
    booktitle={International Conference on Computer Vision (ICCV)},
    year={2021}
}

For a quick summary of this paper, please watch the video presentation.

About

We propose DynAMical Pose estimation (DAMP), the first general and practical framework to perform pose estimation from 2D and 3D visual correspondences by simulating rigid body dynamics arising from virtual springs and damping (top row, magenta lines). DAMP almost always returns the globally optimal rigid transformation across five pose estimation problems (bottom row). (a) Point cloud registration using the Bunny dataset; (b) Primitive registration using a robot model of spheres, planes, cylinders and cones; (c) Category registration using the chair category from the PASCAL3D+ dataset; (d) Absolute pose estimation (APE) using the SPEED satellite dataset; (e) Category APE using the FG3DCar dataset.

Examples

Simply run example_XXX.m files to see the satisfying dynamical pose estimation!

damp's Issues

Hyperparameters tuning and effects?

Hi, thanks for sharing this work.
I just have one question regarding the hyperparameters setting, e.g. time step, damping values.
E.g., if the time step is set to 1 for the absolute pose estimation case, the program will occasionally return Nan results, does this indicate that the settings of those hyperparameters have some influences on the convergence? How should we tune those hyperparameters? Could you give me some insights?

Thanks!