View Selection for 3D Captioning via Diffusion Ranking

arXiv:2404.07984v1 Announce Type: new
Abstract: Scalable annotation approaches are crucial for constructing extensive 3D-text datasets, facilitating a broader range of applications. However, existing methods sometimes lead to the generation of hallucinated captions, compromising caption quality. This paper explores the issue of hallucination in 3D object captioning, with a focus on Cap3D method, which renders 3D objects into 2D views for captioning using pre-trained models. We pinpoint a major challenge: certain rendered views of 3D objects are atypical, deviating from the …

3d object abstract annotation applications arxiv captioning captions cs.cv datasets diffusion focus hallucination however issue object paper quality ranking scalable text type via view