search

wenbowen123 / iros20-6d-pose-tracking

[IROS 2020] se(3)-TrackNet: Data-driven 6D Pose Tracking by Calibrating Image Residuals in Synthetic Domains
Other
388 stars 67 forks source link

Multi-object tracking #4

Closed pyni closed 4 years ago

pyni commented 4 years ago

Nice work! i have two questions: 1) i wonder how you do multi-object tracking shown in your video. 2) why Lie algebra is better than other alternatives to regress the rotation, such as quaternion or this paper[1]?

[1]Zhou, Y., Barnes, C., Lu, J., Yang, J., Li, H.: On the continuity of rotation representations in neural networks. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. (2019) 5745–5753

wenbowen123 commented 4 years ago
  1. For the comparison methods and se(3)-TrackNet, the tracking is applied to each object individually. The estimations are plotted together for visualization only. Sorry for the confusion. If there is a need for multi-object tracking, a naive way is to run multiple networks of se(3)-TrackNet simultaneously.
  2. Good question. This representation has been found empirically numerical unstable for this task. On the other hand, Lie algebra representation is compact and easy to train to get satisfactory results. Thus it is used.
pyni commented 4 years ago

Thanks for replying.