Finding Dino: A plug-and-play framework for unsupervised detection of out-of-distribution objects using prototypes

arXiv:2404.07664v1 Announce Type: new
Abstract: Detecting and localising unknown or Out-of-distribution (OOD) objects in any scene can be a challenging task in vision. Particularly, in safety-critical cases involving autonomous systems like automated vehicles or trains. Supervised anomaly segmentation or open-world object detection models depend on training on exhaustively annotated datasets for every domain and still struggle in distinguishing between background and OOD objects. In this work, we present a plug-and-play generalised framework - PRototype-based zero-shot OOD detection Without Labels (PROWL). …

abstract anomaly arxiv automated automated vehicles autonomous autonomous systems cases cs.ai cs.cv detection distribution framework object objects open-world safety safety-critical segmentation systems training trains type unsupervised vehicles vision world