Bound Tightening Network for Robust Crowd Counting

TL;DR

This paper introduces a Bound Tightening Network (BTN) that enhances the robustness of crowd counting models by propagating interval bounds and utilizing layer weights, improving certified robustness without sacrificing accuracy.

Contribution

The paper proposes a novel BTN architecture that integrates bound propagation and regularization for certified robustness in crowd counting tasks.

Findings

Effective robustness certification demonstrated on benchmark datasets.

Improved counting accuracy with robustness guarantees.

Efficient training process for robust crowd counting models.

Abstract

Crowd Counting is a fundamental topic, aiming to estimate the number of individuals in the crowded images or videos fed from surveillance cameras. Recent works focus on improving counting accuracy, while ignoring the certified robustness of counting models. In this paper, we propose a novel Bound Tightening Network (BTN) for Robust Crowd Counting. It consists of three parts: base model, smooth regularization module and certify bound module. The core idea is to propagate the interval bound through the base model (certify bound module) and utilize the layer weights (smooth regularization module) to guide the network learning. Experiments on different benchmark datasets for counting demonstrate the effectiveness and efficiency of BTN.