Discovering motion hierarchies via tree-structured coding of trajectories

Abstract : The dynamic content of physical scenes is largely compositional, that is, the movements of the objects and of their parts are hierarchically organised and relate through composition along this hierarchy. This structure also prevails in the apparent 2D motion that a video captures. Accessing this visual motion hierarchy is important to get a better understanding of dynamic scenes and is useful for video manipulation. We propose to capture it through learned, tree-structured sparse coding of point trajectories. We leverage this new representation within an unsupervised clustering scheme to partition hierarchically the trajectories into meaningful groups. We show through experiments on motion capture data that our model is able to extract moving segments along with their organisation. We also present competitive results on the task of segmenting objects in video sequences from trajectories.
Document type :
Conference papers
Complete list of metadatas

Cited literature [36 references]  Display  Hide  Download
Contributor : Juan-Manuel Perez-Rua <>
Submitted on : Wednesday, August 31, 2016 - 7:04:40 PM
Last modification on : Wednesday, April 11, 2018 - 1:51:57 AM
Long-term archiving on : Friday, December 2, 2016 - 12:35:32 AM


Files produced by the author(s)


Distributed under a Creative Commons Attribution - NonCommercial - ShareAlike 4.0 International License


  • HAL Id : hal-01358454, version 1



Juan-Manuel Pérez-Rúa, Tomas Crivelli, Patrick Pérez, Patrick Bouthemy. Discovering motion hierarchies via tree-structured coding of trajectories. Proceedings of the British Machine Vision Conference 2016, BMVA, Sep 2016, York, United Kingdom. ⟨hal-01358454⟩



Record views


Files downloads