Represent, Compare, and Learn: A Similarity-Aware Framework for Class-Agnostic Counting

TL;DR

This paper introduces a similarity-aware framework for class-agnostic counting that jointly learns feature representation and similarity metrics, significantly improving counting accuracy and robustness across datasets.

Contribution

It proposes a novel framework that models similarity from multiple aspects and learns it explicitly, advancing the state-of-the-art in class-agnostic counting.

Findings

Our models outperform existing CAC methods on FSC147 dataset.

BMNet+ demonstrates strong cross-dataset generality on CARPK.

Learnable similarity metrics improve counting robustness.

Abstract

Class-agnostic counting (CAC) aims to count all instances in a query image given few exemplars. A standard pipeline is to extract visual features from exemplars and match them with query images to infer object counts. Two essential components in this pipeline are feature representation and similarity metric. Existing methods either adopt a pretrained network to represent features or learn a new one, while applying a naive similarity metric with fixed inner product. We find this paradigm leads to noisy similarity matching and hence harms counting performance. In this work, we propose a similarity-aware CAC framework that jointly learns representation and similarity metric. We first instantiate our framework with a naive baseline called Bilinear Matching Network (BMNet), whose key component is a learnable bilinear similarity metric. To further embody the core of our framework, we extend BMNet to BMNet+ that models similarity from three aspects: 1) representing the instances via their self-similarity to enhance feature robustness against intra-class variations; 2) comparing the similarity dynamically to focus on the key patterns of each exemplar; 3) learning from a supervision signal to impose explicit constraints on matching results. Extensive experiments on a recent CAC dataset FSC147 show that our models significantly outperform state-of-the-art CAC approaches. In addition, we also validate the cross-dataset generality of BMNet and BMNet+ on a car counting dataset CARPK. Code is at tiny.one/BMNet