Identifying the Unknown: Prompt-Free Open Vocabulary Anomaly Recognition for Robot-Object Interaction

2026-06-25

AI Summary

This work addresses the challenge of recognizing unknown objects in open environments without relying on textual prompts or predefined category lists. The authors propose AnomNOVIC, a novel framework that achieves prompt-free, open-vocabulary anomaly detection and classification for the first time. AnomNOVIC integrates a Masked Autoencoder (MAE) with a NOVIC classifier: the MAE generates generic, object-agnostic bounding boxes in an unsupervised manner, while NOVIC performs real-time open-vocabulary classification on salient regions. Evaluated in the NICOL robotic tabletop setting, AnomNOVIC attains 47.1% AP (57.5% AP50); on an in-the-wild test set comprising 48 object categories, it achieves 82.6% accuracy, substantially outperforming baselines such as YOLO-World-v2 and OWLv2. This approach eliminates dependence on candidate class lists and enables continuous deployment in real-world scenarios.

Abstract

Robots operating in real-world environments must in general be able to recognize previously unseen objects. As robotic systems move toward open-world autonomy, there is a growing, yet largely unmet, need for open vocabulary object detectors that are prompt-free and efficient enough for continuous deployment. We present AnomNOVIC, a two-stage known-workspace framework that combines a masked autoencoder (MAE) trained for anomaly detection, with NOVIC, a powerful real-time prompt-free open vocabulary image classifier. The MAE produces generic object-agnostic bounding boxes, allowing NOVIC to classify salient image regions without requiring a predefined candidate class list. We evaluate AnomNOVIC against strong open vocabulary baselines in a tabletop robot-object environment featuring the NICOL humanoid robot, reaching 47.1% AP / 57.5% AP50 for prompt-free recognition, and 59.0% AP / 72.5% AP50 if class candidates are provided. Across additional datasets, including an in-the-wild test set with 48 unique objects, AnomNOVIC reaches up to 82.6% prompt-free detection and classification accuracy. These results significantly surpass all tested open vocabulary baselines, including YOLO-World-v2, OWLv2, and YOLOE.

Problem

• open vocabulary
• anomaly recognition
• prompt-free
• robot-object interaction
• unknown object detection

Innovation

• prompt-free
• open vocabulary
• anomaly detection
• masked autoencoder
• real-time classification