Skip to content
Open access

GAPrompt++: Multi-Granular Geometry-Aware Point Cloud Prompt for 3D Vision Model.

Sep 2026 · IEEE Transactions on Pattern Analysis and Machine Intelligence · Vol PP · 0 citations · 92 references
Medicine Computer Science

TL;DR

GAPrompt++ is proposed, a multi-granular geometry-aware prompting method that provides richer geometric guidance for efficient 3D task adaptation and achieves state-of-the-art performance among prompting-based PEFT methods and even surpasses full fine-tuning across diverse benchmarks, while requiring less than 2% trainable parameters.

Abstract

Pre-trained 3D vision models have substantially advanced point cloud analysis, yet adapting them to downstream tasks via full fine-tuning is computationally expensive and storage-intensive. Parameter-Efficient Fine-Tuning (PEFT) offers a promising alternative by reducing both adaptation cost and storage burden. However, existing prompting-based approaches ignore the intrinsic geometric structures of point clouds, thereby limiting their adaptation capability. This limitation stems from their inability to encode both fine-grained geometric cues and coarse-grained structural semantics, as well as failing to propagate such information effectively through the model hierarchy. To address these challenges, we propose GAPrompt++, a multi-granular geometry-aware prompting method that provides richer geometric guidance for efficient 3D task adaptation. Specifically, we introduce a Point Shift Prompter that extracts multi-granular geometric features across different scales, enabling instance-specific geometric adjustments during adaptation. Next, a Keypoint Prompter adaptively generates point-level prompts to highlight local geometric saliency and fine-grained structural details. Furthermore, a Prompt Propagation mechanism injects these multi-granular geometric cues throughout the feature extraction hierarchy, strengthening the ability to capture essential geometric characteristics. Extensive experiments show that GAPrompt++ achieves state-of-the-art performance among prompting-based PEFT methods and even surpasses full fine-tuning across diverse benchmarks, while requiring less than 2% trainable parameters. In addition, to address the saturation of existing evaluation datasets, we construct two more challenging benchmarks derived from 3D Gaussian Splatting and Multi-View Stereo reconstruction, offering diverse and realistic point cloud scenarios to promote future research.

Read PDF

Similar papers

Sep 2026

Fine-grained pillar encoder with spatial adaptation for lidar-based 3D object detection

Point clouds based object detection is a fundamental and challenge task in three-dimensional computer vision. However, existing point-based methods struggle to meet real-time object recognition requirements due to excessive raw point counts, whereas pillar-based approaches are constrained by point cloud aggregation alg...

Unknown authors · 0 citations
Preprint Sep 2026

Partition-Invariant Tuning for 3D Scene Understanding

PointPiT is proposed, a partition-invariant tuning framework for scene-level point clouds that integrates local geometric patterns with global scene context to mitigate partition-induced representation shifts, and achieves consistent state-of-the-art performance among representative PEFT methods.

Hong-Qiang Lin, Tian-Le Wang, Shui-Wang Li et al. · 0 citations
Oct 2026

MLCompletion: Point Cloud Completion via Multiview Feature Augmentation and Progressive Refinement

Point cloud completion is a fundamental task in 3-D sensing and vision, yet it remains challenging when partial observations suffer from severe structural missing regions and ambiguous local geometry. To address these issues, we propose MLCompletion, a point-cloud-only completion framework that incorporates self-genera...

Zhi-Qiang Liu, Lin-Bo Hao, Shao-Hua Jiang et al. · 0 citations
Conference Open access Sep 2026

PointGP: Geometry-Primed Attention for Point Cloud Analysis

PointGP is proposed, a geometry-primed framework that uses rectified local geometric topology as the primary cue for attention generation and achieves competitive accuracy with strong parameter and computational efficiency compared with representative strong baselines.

Yong Yang, Jian-Min Huang, Meng-Yuan Ge et al. · 0 citations
Preprint Aug 2026

Stream3Dv2: Geometric-Semantic Fusion Enhanced Streaming Zero-Shot 3D Scene Understanding

A comprehensive geometric-semantic fusion mechanism that resolves geometric noise and semantic ambiguity by explicitly utilizing semantic guidance and formulating 3D segmentation as solving point-and-set merging and partitioning problems, and an innovative manifold-distance-based point cloud refinement strategy.

Jie Xu, Na Zhao · 0 citations
Open access Aug 2026

Geometry-Enhanced Point–Voxel Fusion with Active Learning for Label-Efficient Point Cloud Semantic Segmentation

GeoFuse-AL is proposed, a label-efficient segmentation framework that integrates a geometry-guided point–voxel network with multi-cue active learning and improves segmentation accuracy while substantially reducing annotation requirements.

Cheng Zhang, Fei Meng, Yi-Chang Qiu et al. · 0 citations

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.