TransRefer3D: Entity-and-Relation Aware Transformer for Fine-Grained 3D Visual Grounding
This work addresses the challenge of distinguishing similar objects in 3D scenes for applications like robotics and augmented reality, presenting a novel method that advances the field beyond previous graph-based approaches.
The paper tackles the problem of fine-grained 3D visual grounding, which involves identifying a specific 3D object from distractors of the same category using natural language, by proposing a Transformer-based network called TransRefer3D that uses entity-and-relation aware attention modules for multimodal feature matching. The result is a significant performance improvement, outperforming existing approaches by up to 10.6% on standard datasets and achieving state-of-the-art.
Recently proposed fine-grained 3D visual grounding is an essential and challenging task, whose goal is to identify the 3D object referred by a natural language sentence from other distractive objects of the same category. Existing works usually adopt dynamic graph networks to indirectly model the intra/inter-modal interactions, making the model difficult to distinguish the referred object from distractors due to the monolithic representations of visual and linguistic contents. In this work, we exploit Transformer for its natural suitability on permutation-invariant 3D point clouds data and propose a TransRefer3D network to extract entity-and-relation aware multimodal context among objects for more discriminative feature learning. Concretely, we devise an Entity-aware Attention (EA) module and a Relation-aware Attention (RA) module to conduct fine-grained cross-modal feature matching. Facilitated by co-attention operation, our EA module matches visual entity features with linguistic entity features while RA module matches pair-wise visual relation features with linguistic relation features, respectively. We further integrate EA and RA modules into an Entity-and-Relation aware Contextual Block (ERCB) and stack several ERCBs to form our TransRefer3D for hierarchical multimodal context modeling. Extensive experiments on both Nr3D and Sr3D datasets demonstrate that our proposed model significantly outperforms existing approaches by up to 10.6% and claims the new state-of-the-art. To the best of our knowledge, this is the first work investigating Transformer architecture for fine-grained 3D visual grounding task.