NAS-Count: Counting-by-Density with Neural Architecture Search

TL;DR

This paper introduces NAS-Count, an automated neural architecture search method for crowd counting, resulting in an end-to-end encoder-decoder model that outperforms existing hand-designed models on multiple datasets.

Contribution

The paper presents AMSNet, a novel NAS-designed crowd counting architecture with a counting-specific search space and a new scale pyramid pooling loss, advancing automation in model design.

Findings

AMSNet achieves state-of-the-art results on four datasets.

The NAS approach outperforms hand-designed models.

The scale pyramid pooling loss improves multi-scale feature learning.

Abstract

Most of the recent advances in crowd counting have evolved from hand-designed density estimation networks, where multi-scale features are leveraged to address the scale variation problem, but at the expense of demanding design efforts. In this work, we automate the design of counting models with Neural Architecture Search (NAS) and introduce an end-to-end searched encoder-decoder architecture, Automatic Multi-Scale Network (AMSNet). Specifically, we utilize a counting-specific two-level search space. The encoder and decoder in AMSNet are composed of different cells discovered from micro-level search, while the multi-path architecture is explored through macro-level search. To solve the pixel-level isolation issue in MSE loss, AMSNet is optimized with an auto-searched Scale Pyramid Pooling Loss (SPPLoss) that supervises the multi-scale structural information. Extensive experiments on four datasets show AMSNet produces state-of-the-art results that outperform hand-designed models, fully demonstrating the efficacy of NAS-Count.