Learning Spatial-Aware Regressions for Visual Tracking

TL;DR

This paper introduces a novel visual tracking method combining a kernelized ridge regression model and a spatially regularized convolutional neural network to improve robustness by leveraging spatial information of deep features.

Contribution

It proposes a dual regression framework that integrates kernelized ridge regression with a spatially regularized CNN for enhanced visual tracking performance.

Findings

Effective in handling target appearance variations

Outperforms existing methods on benchmark datasets

Robust against occlusion and background clutter

Abstract

In this paper, we analyze the spatial information of deep features, and propose two complementary regressions for robust visual tracking. First, we propose a kernelized ridge regression model wherein the kernel value is defined as the weighted sum of similarity scores of all pairs of patches between two samples. We show that this model can be formulated as a neural network and thus can be efficiently solved. Second, we propose a fully convolutional neural network with spatially regularized kernels, through which the filter kernel corresponding to each output channel is forced to focus on a specific region of the target. Distance transform pooling is further exploited to determine the effectiveness of each output channel of the convolution layer. The outputs from the kernelized ridge regression model and the fully convolutional neural network are combined to obtain the ultimate response. Experimental results on two benchmark datasets validate the effectiveness of the proposed method.