Newest papers

Open-Vocabulary Models and Label Propagation for Uncertainty-aware Pseudo-Labeling of Object Detection Datasets

Jonathan Klees, Timo Heller, Matthias Rottmann · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.29788 · cited by 0 works (OpenAlex) · open-access PDF

Korean

객체 감지 효율적인 모델은 대규모 데이터 인코딩 데이터 세트에 의존하여 상당한 수동 인코딩 노력을 필요로합니다. 인코딩 비용은 인코딩 불확실성이 부드러운 라벨을 통해 캡처되면 더욱 증가합니다. 각 개체에 대해 다수의 인코딩이 수집되어야합니다. 우리는 이 도전을 개체 수준의 유사성 그래프를 구축하고 확률적 인 라벨 정보를 전파함으로써 상당 규모의 데이터 세트의 불확실성 인코딩에 대한 프레임 워크를 도입합니다. 우리의 접근 방식은 개체 감지에 대한 지역 제안과 상징적으로 의미있는 개체 표현을 생성하기 위해 오픈 사전적 인코딩 객체 탐지기를 활용합니다. 이러한 표현에 기초하여 우리는 객체 수준의 유

Machine translation (Meta M2M100); the English original follows.

Abstract

Object detection models rely on large-scale annotated datasets that require substantial manual annotation effort. The labeling costs further increase when annotation uncertainty is captured through soft labels, as multiple annotations must be collected for each object. We address this challenge by introducing a semi-supervised framework for the uncertainty-aware annotation of object detection datasets. Our approach leverages open-vocabulary object detectors to generate region proposals and semantically meaningful object representations. Based on these representations, we construct an object-level similarity graph and propagate probabilistic label information. Our framework assigns soft labels to previously unlabeled objects, quantifying both aleatoric and epistemic uncertainty. We evaluate our method on four object detection benchmarks with soft-label annotations and demonstrate that high-quality annotations can be obtained with only a small fraction of manually labeled objects. Furthermore, we show that detectors trained on semi-automatically annotated data achieve performance comparable to models trained on fully manually annotated data. We achieve state-of-the-art downstream detection performance for the task of semi-supervised object detection on the COCO dataset, with 40.5% AP virtually matching the performance of the model trained on fully manually constructed data (40.9%) at only 1% of labeling costs. These results highlight the potential of open-vocabulary models and graph-based label propagation for efficient, accurate and uncertainty-aware pseudo-labeling.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Annotation, Computer science, Object detection, Artificial intelligence, Probabilistic logic, Object (grammar)

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange