Introduces a Prompt-Free Universal Region Proposal Network (PF-RPN) that identifies objects in any domain without needing text or image exemplars.
arXiv · March 19, 2026 · 2603.17554
The Takeaway
It enables high-quality object localization across diverse domains (underwater, industrial, remote sensing) with zero fine-tuning and minimal training data. This is a significant leap for autonomous systems and 'open-world' vision tasks where categories are not known in advance.
From the abstract
Identifying potential objects is critical for object recognition and analysis across various computer vision applications. Existing methods typically localize potential objects by relying on exemplar images, predefined categories, or textual descriptions. However, their reliance on image and text prompts often limits flexibility, restricting adaptability in real-world scenarios. In this paper, we introduce a novel Prompt-Free Universal Region Proposal Network (PF-RPN), which identifies potential