SenseFuse improves 3D object recognition by combining image and shape data
SenseFuse: Label-Free Fusion of Image and Shape Encoders for Open-Vocabulary 3D Instance Segmentation
Computer Vision and Pattern RecognitionRobotics
Summary
Identifying objects in 3D scenes is important for robots to understand and interact with their surroundings. Existing methods that recognize many types of objects usually rely only on 2D images, missing out on useful 3D shape information. The authors introduced SenseFuse, a technique that blends 2D image features and 3D shape data without needing extra labels, improving how well objects are detected in 3D scenes. Their method adapts to each scene quickly and works better than previous approaches in many tests.
What this means in practice
- •For robotics developers: Enable robots to better identify and segment objects in 3D environments using both image and shape data for improved manipulation tasks.
- •For 3d mapping engineers: Enhance 3D scene mapping by integrating image and geometric cues to label objects without needing manual annotations.
Authors
Euiseok Han, Tri Ton, Hwanhee Kim, Seungyeon Ryu, Chang D. Yoo
Abstract
Open-vocabulary scene understanding is fundamental for robotics, laying the groundwork for spatial reasoning and object manipulation. While closed-vocabulary 3D instance segmentation heavily leverages 3D shape information, state-of-the-art open-vocabulary methods remain predominantly restricted to 2D image features or image-distilled representations during mask labeling. In this paper, we propose SenseFuse, a label-free fusion method that balances 2D image and 3D shape encoders for robust open-vocabulary 3D instance segmentation, refining only the mask-labeling stage of existing pipelines. We reveal that 2D image and 3D shape encoders exhibit largely disjoint failure patterns and rarely share identical wrong labels, whereas two 2D image encoders frequently repeat the same errors. This distinct behavior makes the 2D and 3D pair inherently complementary. We introduce an adaptive mechanism that selects a scene-level fusion weight to maximize a label-free sensitivity measure, estimated directly from a single scene's unlabeled proposals in milliseconds. SenseFuse improves labeling accuracy in every evaluated setting across ScanNet200, Replica, and ScanNet++, recovering 67-100% (median 93%) of the gain achievable with an oracle weight, and it raises instance AP in 21 of 22 reported settings. Code is available at https://github.com/hanes1207/SenseFuse.