Bound Tightening Network for Robust Crowd Counting
This work addresses the need for robust crowd counting in surveillance applications, but it appears incremental as it builds on existing accuracy-focused methods by adding robustness certification.
The paper tackles the problem of improving certified robustness in crowd counting models, which previous works had ignored, and demonstrates the effectiveness of their proposed Bound Tightening Network (BTN) through experiments on benchmark datasets.
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.