Bridging the Gap between Object and Image-level Representations for Open-Vocabulary Detection

TL;DR

This paper introduces a novel approach to open-vocabulary detection that aligns object and image-level representations using object-centric alignment of CLIP embeddings and pseudo-labeling, significantly improving detection performance on COCO and LVIS benchmarks.

Contribution

It proposes a new method combining object-centric alignment of CLIP embeddings with pseudo-labeling for image-level supervision, bridging the gap between object and image representations in open-vocabulary detection.

Findings

Achieves 36.6 AP50 on COCO novel classes, 8.2 points higher than previous best.

Surpasses state-of-the-art ViLD model by 5.0 mask AP on LVIS rare categories.

Effectively reduces the gap between object and image-centric representations.

Abstract

Existing open-vocabulary object detectors typically enlarge their vocabulary sizes by leveraging different forms of weak supervision. This helps generalize to novel objects at inference. Two popular forms of weak-supervision used in open-vocabulary detection (OVD) include pretrained CLIP model and image-level supervision. We note that both these modes of supervision are not optimally aligned for the detection task: CLIP is trained with image-text pairs and lacks precise localization of objects while the image-level supervision has been used with heuristics that do not accurately specify local object regions. In this work, we propose to address this problem by performing object-centric alignment of the language embeddings from the CLIP model. Furthermore, we visually ground the objects with only image-level supervision using a pseudo-labeling process that provides high-quality object proposals and helps expand the vocabulary during training. We establish a bridge between the above two object-alignment strategies via a novel weight transfer function that aggregates their complimentary strengths. In essence, the proposed model seeks to minimize the gap between object and image-centric representations in the OVD setting. On the COCO benchmark, our proposed approach achieves 36.6 AP50 on novel classes, an absolute 8.2 gain over the previous best performance. For LVIS, we surpass the state-of-the-art ViLD model by 5.0 mask AP for rare categories and 3.4 overall. Code: https://github.com/hanoonaR/object-centric-ovd.