Pith. sign in

REVIEW 12 cited by

Swin3D: A Pretrained Transformer Backbone for 3D Indoor Scene Understanding

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2304.06906 v3 pith:6MUI2EJ3 submitted 2023-04-14 cs.CV

Yu-Qi Yang , Yu-Xiao Guo , Jian-Yu Xiong , Yang Liu , Hao Pan , Peng-Shuai Wang , Xin Tong , Baining Guo This is my paper
classification cs.CV
keywords pretrainedbackbonedatasetdetectionmiouscannetsegmentationdatasets
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

The use of pretrained backbones with fine-tuning has been successful for 2D vision and natural language processing tasks, showing advantages over task-specific networks. In this work, we introduce a pretrained 3D backbone, called {\SST}, for 3D indoor scene understanding. We design a 3D Swin transformer as our backbone network, which enables efficient self-attention on sparse voxels with linear memory complexity, making the backbone scalable to large models and datasets. We also introduce a generalized contextual relative positional embedding scheme to capture various irregularities of point signals for improved network performance. We pretrained a large {\SST} model on a synthetic Structured3D dataset, which is an order of magnitude larger than the ScanNet dataset. Our model pretrained on the synthetic dataset not only generalizes well to downstream segmentation and detection on real 3D point datasets, but also outperforms state-of-the-art methods on downstream tasks with +2.3 mIoU and +2.2 mIoU on S3DIS Area5 and 6-fold semantic segmentation, +1.8 mIoU on ScanNet segmentation (val), +1.9 mAP@0.5 on ScanNet detection, and +8.1 mAP@0.5 on S3DIS detection. A series of extensive ablation studies further validate the scalability, generality, and superior performance enabled by our approach. The code and models are available at https://github.com/microsoft/Swin3D .

Discussion (0). Sign in to comment.

Forward citations

Cited by 12 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Generalized Zero-Shot Learning for Point Cloud Segmentation with Evidence-Based Dynamic Calibration

    cs.CV 2025-09 conditional novelty 6.0 of 10

    E3DPC-GZSL dynamically calibrates seen-class probabilities using an evidence-based uncertainty estimate, improving generalized zero-shot 3D segmentation over fixed-calibration baselines.

  2. FastPoint: Accelerating 3D Point Cloud Model Inference via Sample Point Distance Prediction

    cs.CV 2025-07 conditional novelty 6.0 of 10

    FastPoint accelerates farthest point sampling and neighbor search by predicting the FPS minimum distance curve from its first 10%, achieving 2.55x end-to-end speedup with negligible accuracy loss.

  3. Asymmetric Dual Self-Distillation for 3D Self-Supervised Representation Learning

    cs.CV 2025-06 reject novelty 6.0 of 10

    AsymDSD unifies latent masked point modeling and cross-view invariance self-distillation to learn 3D representations, reporting 90.53% on ScanObjectNN and 93.72% with 930k-shape pretraining.

  4. UniPre3D: Unified Pre-training of 3D Point Cloud Models with Cross-Modal Gaussian Splatting

    cs.CV 2025-06 conditional novelty 6.0 of 10

    A point cloud pre-training method that uses 3D Gaussian splatting rendering and cross-modal image features to work for both objects and scenes.

  5. Towards Explicit Geometry-Reflectance Collaboration for Generalized LiDAR Segmentation in Adverse Weather

    cs.CV 2025-06 conditional novelty 6.0 of 10

    A dual-branch geometry and reflectance network with information-bottleneck and multi-level fusion improves LiDAR semantic segmentation generalization under fog, rain, and snow.

  6. TCFNet: Bidirectional face-bone transformation via a Transformer-based coarse-to-fine point movement network

    cs.CV 2025-08 conditional novelty 5.0 of 10

    TCFNet is a two-stage Transformer-based network that moves face and bone point clouds to predict facial shape changes for jaw surgery, reporting better accuracy than prior deep learning methods on gathered CT datasets.

  7. Structured Spectral Graph Learning for Anomaly Classification in 3D Chest CT Scans

    cs.CV 2025-08 conditional novelty 5.0 of 10

    A graph neural network with spectral convolution over axial slice triplet nodes improves multi-label chest CT anomaly classification on two public datasets.

  8. HydraMamba: Multi-Head State Space Model for Global Point Cloud Learning

    cs.CV 2025-07 conditional novelty 5.0 of 10

    A state space model based point cloud network with shuffled Hilbert serialization, a convolutional bidirectional S6 branch, and multi-head S6 achieves new top scores on ModelNet40, ShapeNet, S3DIS, and ScanObjectNN.

  9. OmniVec2 -- A Novel Transformer based Network for Large Scale Multimodal and Multitask Learning

    cs.CV 2025-07 conditional novelty 5.0 of 10

    A shared-backbone transformer with pairwise modality training reports top results across 25 datasets spanning 12 modalities.

  10. Towards Generalized Range-View LiDAR Segmentation in Adverse Weather

    cs.CV 2025-06 conditional novelty 5.0 of 10

    A plug-in stem-block framework with geometric noise suppression and reflectance calibration boosts range-view LiDAR segmentation accuracy in adverse weather by large margins.

  11. Structured Spectral Graph Representation Learning for Multi-label Abnormality Analysis from 3D CT Scans

    cs.CV 2025-10 conditional novelty 4.0 of 10

    A graph-of-slice-triplets encoder with spectral convolution outperforms 3D CNN/Transformer baselines on multi-label chest CT abnormality classification and transfers to report generation and abdominal CT.

  12. HybridTM: Combining Transformer and Mamba for 3D Semantic Segmentation

    cs.CV 2025-07 conditional novelty 4.0 of 10

    A new 3D semantic segmentation architecture that interleaves attention and Mamba operators within each layer achieves small but consistent gains on indoor and outdoor benchmarks.

Pith tools