Toward Ambulatory Vision: Learning Visually-Grounded Active View Selection
This addresses the problem of enabling embodied agents to perform ambulatory vision for improved visual reasoning, representing an incremental advancement in active vision tasks.
The paper tackles the limitation of Vision Language Models (VLMs) to static images by introducing Visually Grounded Active View Selection (VG-AVS), a task for selecting informative next viewpoints using only current visual information, and achieves strong question answering performance with generalization to unseen scenes.
Vision Language Models (VLMs) excel at visual question answering (VQA) but remain limited to snapshot vision, reasoning from static images. In contrast, embodied agents require ambulatory vision, actively moving to obtain more informative views. We introduce Visually Grounded Active View Selection (VG-AVS), a task that selects the most informative next viewpoint using only the visual information in the current image, without relying on scene memory or external knowledge. To support this task, we construct a synthetic dataset with automatically generated paired query-target views and question-answer prompts. We also propose a framework that fine-tunes pretrained VLMs through supervised fine-tuning (SFT) followed by RL-based policy optimization. Our approach achieves strong question answering performance based on viewpoint selection and generalizes robustly to unseen synthetic and real scenes. Furthermore, incorporating our learned VG-AVS framework into existing scene-exploration-based EQA systems improves downstream question-answering accuracy.