UN-DETR: Promoting Objectness Learning via Joint Supervision for Unknown Object Detection

TL;DR

UN-DETR introduces a transformer-based framework for unknown object detection that leverages joint supervision of objectness, improving detection of unseen categories with state-of-the-art results.

Contribution

The paper proposes UN-DETR, a novel transformer-based UOD framework that uses joint supervised learning of objectness from positional and categorical features, along with new strategies for assignment and filtering.

Findings

Achieves state-of-the-art performance on UOD benchmarks.

Effectively detects unknown objects with improved accuracy.

Demonstrates the benefit of joint supervision for objectness learning.

Abstract

Unknown Object Detection (UOD) aims to identify objects of unseen categories, differing from the traditional detection paradigm limited by the closed-world assumption. A key component of UOD is learning a generalized representation, i.e. objectness for both known and unknown categories to distinguish and localize objects from the background in a class-agnostic manner. However, previous methods obtain supervision signals for learning objectness in isolation from either localization or classification information, leading to poor performance for UOD. To address this issue, we propose a transformer-based UOD framework, UN-DETR. Based on this, we craft Instance Presence Score (IPS) to represent the probability of an object's presence. For the purpose of information complementarity, IPS employs a strategy of joint supervised learning, integrating attributes representing general objectness from the positional and the categorical latent space as supervision signals. To enhance IPS learning, we introduce a one-to-many assignment strategy to incorporate more supervision. Then, we propose Unbiased Query Selection to provide premium initial query vectors for the decoder. Additionally, we propose an IPS-guided post-process strategy to filter redundant boxes and correct classification predictions for known and unknown objects. Finally, we pretrain the entire UN-DETR in an unsupervised manner, in order to obtain objectness prior. Our UN-DETR is comprehensively evaluated on multiple UOD and known detection benchmarks, demonstrating its effectiveness and achieving state-of-the-art performance.