Hierarchical Few-Shot Object Detection: Problem, Benchmark and Method

TL;DR

This paper introduces hierarchical few-shot object detection (Hi-FSOD), presents a large-scale bird dataset with taxonomy, and proposes a hierarchical contrastive learning method that improves detection accuracy in hierarchical categories.

Contribution

It defines the new Hi-FSOD problem, creates the first large-scale Hi-FSOD benchmark dataset HiFSOD-Bird, and proposes the HiCLPL method utilizing hierarchical contrastive learning and probabilistic loss.

Findings

HiCLPL outperforms existing FSOD methods on HiFSOD-Bird.

The dataset contains 176,350 images with 1,432 hierarchical categories.

Hierarchical contrastive learning improves feature space organization.

Abstract

Few-shot object detection (FSOD) is to detect objects with a few examples. However, existing FSOD methods do not consider hierarchical fine-grained category structures of objects that exist widely in real life. For example, animals are taxonomically classified into orders, families, genera and species etc. In this paper, we propose and solve a new problem called hierarchical few-shot object detection (Hi-FSOD), which aims to detect objects with hierarchical categories in the FSOD paradigm. To this end, on the one hand, we build the first large-scale and high-quality Hi-FSOD benchmark dataset HiFSOD-Bird, which contains 176,350 wild-bird images falling to 1,432 categories. All the categories are organized into a 4-level taxonomy, consisting of 32 orders, 132 families, 572 genera and 1,432 species. On the other hand, we propose the first Hi-FSOD method HiCLPL, where a hierarchical contrastive learning approach is developed to constrain the feature space so that the feature distribution of objects is consistent with the hierarchical taxonomy and the model's generalization power is strengthened. Meanwhile, a probabilistic loss is designed to enable the child nodes to correct the classification errors of their parent nodes in the taxonomy. Extensive experiments on the benchmark dataset HiFSOD-Bird show that our method HiCLPL outperforms the existing FSOD methods.