Bi-level Alignment for Cross-Domain Crowd Counting

TL;DR

This paper introduces a simple, efficient bi-level adversarial alignment framework for unsupervised domain adaptation in crowd counting, leveraging AutoML for optimal source data transformation and separate foreground-background alignment to improve performance on real-world datasets.

Contribution

The paper proposes a novel bi-level alignment framework combining AutoML-based data transformation and fine-grained feature alignment for unsupervised domain adaptation in crowd counting.

Findings

Outperforms existing methods on five crowd counting benchmarks.

Achieves large margin improvements with a simple and efficient approach.

Demonstrates the effectiveness of bi-level alignment and AutoML in domain adaptation.

Abstract

Recently, crowd density estimation has received increasing attention. The main challenge for this task is to achieve high-quality manual annotations on a large amount of training data. To avoid reliance on such annotations, previous works apply unsupervised domain adaptation (UDA) techniques by transferring knowledge learned from easily accessible synthetic data to real-world datasets. However, current state-of-the-art methods either rely on external data for training an auxiliary task or apply an expensive coarse-to-fine estimation. In this work, we aim to develop a new adversarial learning based method, which is simple and efficient to apply. To reduce the domain gap between the synthetic and real data, we design a bi-level alignment framework (BLA) consisting of (1) task-driven data alignment and (2) fine-grained feature alignment. In contrast to previous domain augmentation methods, we introduce AutoML to search for an optimal transform on source, which well serves for the downstream task. On the other hand, we do fine-grained alignment for foreground and background separately to alleviate the alignment difficulty. We evaluate our approach on five real-world crowd counting benchmarks, where we outperform existing approaches by a large margin. Also, our approach is simple, easy to implement and efficient to apply. The code is publicly available at https://github.com/Yankeegsj/BLA.