A Few-Shot Sequential Approach for Object Counting

TL;DR

This paper introduces a novel few-shot multi-class object counting method using class-agnostic attention and prototypical networks, trained with point-level annotations, and demonstrates its effectiveness across multiple datasets including a new weakly supervised dataset.

Contribution

The work proposes a new sequential attention-based approach for few-shot object counting that leverages class-agnostic features and a novel loss function, along with a new dataset for weakly supervised counting.

Findings

Effective on FSOD and MS COCO datasets.

Robust performance on different class distributions.

Introduces a new dataset for weakly supervised counting.

Abstract

In this work, we address the problem of few-shot multi-class object counting with point-level annotations. The proposed technique leverages a class agnostic attention mechanism that sequentially attends to objects in the image and extracts their relevant features. This process is employed on an adapted prototypical-based few-shot approach that uses the extracted features to classify each one either as one of the classes present in the support set images or as background. The proposed technique is trained on point-level annotations and uses a novel loss function that disentangles class-dependent and class-agnostic aspects of the model to help with the task of few-shot object counting. We present our results on a variety of object-counting/detection datasets, including FSOD and MS COCO. In addition, we introduce a new dataset that is specifically designed for weakly supervised multi-class object counting/detection and contains considerably different classes and distribution of number of classes/instances per image compared to the existing datasets. We demonstrate the robustness of our approach by testing our system on a totally different distribution of classes from what it has been trained on.