Deeply optimized Hough transform: Application to action segmentation
Abstract
Hough-like methods like Implicit Shape Model (ISM) and Hough forest have been successfully applied in multiple computer vision fields like object detection, tracking, skeleton extraction or human action detection. However, these methods are known to generate false positives. To handle this issue, several works like Max-Margin Hough Transform (MMHT) or Implicit Shape Kernel (ISK) have reported significant performance improvements by adding discriminative parameters to the generative ones introduced by ISM. In this paper, we offer to use only discriminative parameters that are globally optimized according to all the variables of the Hough transform. To this end, we abstract the common vote process of all Hough methods into linear equations, leading to a training formulation that can be solved using linear programming solvers. Our new Hough Transform significantly outperforms the previous ones on HoneyBee and TUM datasets, two public databases of action and behaviour segmentation.
Origin : Files produced by the author(s)
Loading...