Tracking-by-Segmentation with Online Gradient Boosting Decision Tree

We propose an online tracking algorithm that adaptively models target appearances based on an online gradient boosting decision tree. Our algorithm is particularly use-ful for non-rigid and/or articulated objects since it handles various deformations of the target effectively by integrating a classifier operating on individual patches and provides segmentation masks of the target as final results. The pos-terior of the target state is propagated over time by particle filtering, where the likelihood is computed based mainly on a patch-level confidence map associated with a latent target state corresponding to each sample. Once tracking is com-pleted in each frame, our gradient boosting decision tree is updated to adapt new data in a recursive manner. For effective evaluation of segmentation-based tracking algo-rithms, we construct a new ground-truth that contains pixel-level annotation of segmentation mask. We evaluate the performance of our tracking algorithm based on the mea-sures for segmentation masks, where our algorithm illus-trates outstanding accuracy compared to the state-of-the-art segmentation-based tracking methods. 1