VU-RED-F： improved CAD model replacement for U-RED single-view point clouds

doi:10.11772/j.issn.1001-9081.2025050575

Journal of Computer Applications ›› 2026, Vol. 46 ›› Issue (5): 1534-1544.DOI: 10.11772/j.issn.1001-9081.2025050575

• Multimedia computing and computer simulation • Previous Articles

VU-RED-F： improved CAD model replacement for U-RED single-view point clouds

Gengxin FAN¹^,²^,³, Huiyan HAN¹^,²^,³(), Liqun KUANG¹^,²^,³, Ziyang JIN¹^,²^,³, Huafeng ZHAO¹^,²^,³

^1.School of Computer Science and Technology，North University of China，Taiyuan Shanxi 030051，China
^2.Shanxi Key Laboratory of Machine Vision and Virtual Reality （North University of China），Taiyuan Shanxi 030051，China
^3.Shanxi Provincial Vision Information Processing and Intelligent Robot Engineering Research Center，Taiyuan Shanxi 030051，China

Received:2025-05-27 Revised:2025-09-26 Accepted:2025-10-10 Online:2025-10-17 Published:2026-05-10
Contact: Huiyan HAN
About author:FAN Gengxin， born in 1999， M. S. candidate. Her research interests include point cloud reconstruction.
KUANG Liqun， born in 1976， Ph. D.， professor. His research interests include artificial intelligence， computer vision.
JIN Ziyang， born in 2000， M. S. candidate. Her research interests include 3D reconstruction.
ZHAO Huafeng， born in 2001， M. S. candidate. His research interests include computer vision.
Supported by:
Shanxi Province Natural Science Foundation(202303021211153);Shanxi Province Major Science and Technology Plan “Unveiling the List and Assigning the Leader” Project(202201150401021);Shanxi Province Graduate Research Innovation Project(2024SJ279)

VU-RED-F：改进U-RED的单视角点云CAD模型替换

樊耿鑫¹^,²^,³, 韩慧妍¹^,²^,³(), 况立群¹^,²^,³, 晋紫阳¹^,²^,³, 赵华峰¹^,²^,³

^1.中北大学计算机科学与技术学院，太原 030051
^2.机器视觉与虚拟现实山西省重点实验室（中北大学），太原 030051
^3.山西省视觉信息处理及智能机器人工程研究中心，太原 030051

通讯作者: 韩慧妍
作者简介:樊耿鑫（1999—），女，山西运城人，硕士研究生，CCF会员，主要研究方向：点云重建
况立群（1976—），男，江西高安人，教授，博士，CCF会员，主要研究方向：人工智能、计算机视觉
晋紫阳（2000—），女，山西霍州人，硕士研究生，CCF会员，主要研究方向：三维重建
赵华峰（2001—），男，山西运城人，硕士研究生，CCF会员，主要研究方向：计算机视觉。
基金资助:
山西省自然科学基金资助项目(202303021211153);山西省科技重大专项计划“揭榜挂帅”项目(202201150401021);山西省研究生科研创新项目(2024SJ279)

Abstract

Abstract:

In robotic environmental perception tasks， single-view point clouds suffer from severe geometric information loss due to sensor viewpoint limitations. Point cloud reconstruction methods based on Computer-Aided Design （CAD） model replacement can avoid the risks of structural instability associated with reconstruction directly from point clouds by retrieving similar models and applying deformation. Unsupervised 3D shape REtrieval and Deformation （U-RED） algorithm achieves topologically consistent CAD model replacement while maintaining the editability of the reconstruction results. However， when dealing with objects with complex topology， it still faces problems such as insufficient rotation and translation invariance in point cloud representations， difficulty in distinguishing neighboring components due to geometric similarity among homologous components， and parameter update failures caused by scattered attention weights and gradient vanishing or explosion. To address these challenges， a Vector neuron enhanced Unsupervised REtrieval and Deformation algorithm with Feature affine residual （VU-RED-F） was proposed based on U-RED. Firstly， a Vector Neuron Encoder （VNE） was constructed to improve the robustness of the feature extraction module in representing rotation and translation invariance of point clouds. Secondly， learnable affine transformation residuals were introduced to reconstruct the feature mapping process， adaptively adjust the feature distribution， and enhance the network's ability to discriminate local geometric structures between components. Finally， by integrating soft-threshold gating and residual correction， the stability of gradient propagation was enhanced while constraining the sparsity of the attention distribution， thereby boosting network convergence efficiency and reducing loss during retrieval and deformation. Experimental results on the synthetic PartNet and ComplementMe datasets， as well as the real Scan2CAD dataset， show that the VU-RED-F algorithm has the lowest average chamfer distance （cd） loss， improving the fidelity of local geometric details in CAD models.

Key words: point cloud reconstruction, model retrieval and deformation, vector neuron, Feature Affine Residual (FA?Res), residual network

摘要：

在机器人环境感知任务中，单视角点云因传感器视角受限导致几何信息严重缺失。基于计算机辅助设计（CAD）模型替换的点云重建方法通过检索相似模型并实施变形，可有效规避直接从点云中重建的结构失控风险。无监督三维形状检索与变形（U-RED）算法在保持重建结果可编辑性的同时可实现拓扑一致的CAD模型替换，但面对复杂拓扑结构物体时仍存在点云旋转平移不变性表征不足、同源部件间的几何相似性导致近邻部件区分困难、注意力权重分散以及梯度消失与爆炸引起的参数更新失效问题。针对上述挑战，在U-RED基础上提出基于向量神经元与特征仿射残差增强的无监督检索变形算法（VU-RED-F）。首先，构建向量神经元编码器（VNE），提升特征提取模块在点云旋转平移不变性表征的鲁棒性；其次，引入可学习的仿射变换残差以重构特征映射过程，自适应调整特征分布，增强网络对部件间局部几何结构的判别能力；最后，融合软阈值门控与残差校正，在约束注意力分布稀疏性的同时增强梯度传播稳定性，提升网络收敛效率，降低检索变形过程中的损失。在PartNet和ComplementMe合成数据集以及Scan2CAD真实数据集上的实验结果表明，VU-RED-F算法的平均倒角距离（cd）损失最小，提高了CAD模型的局部几何细节保真度。

关键词: 点云重建, 模型检索与变形, 向量神经元, 特征仿射残差, 残差网络

CLC Number:

TP391.41

Gengxin FAN, Huiyan HAN, Liqun KUANG, Ziyang JIN, Huafeng ZHAO. VU-RED-F： improved CAD model replacement for U-RED single-view point clouds[J]. Journal of Computer Applications, 2026, 46(5): 1534-1544.

樊耿鑫, 韩慧妍, 况立群, 晋紫阳, 赵华峰. VU-RED-F：改进U-RED的单视角点云CAD模型替换[J]. 《计算机应用》唯一官方网站, 2026, 46(5): 1534-1544.

Figures/Tables 17

Fig. 1 Framework of U-RED algorithm

Fig. 2 Framework of VU-RED-F algorithm

Fig. 3 Structure of VNE

Fig. 4 Structure of FA-Res

Fig. 5 Structure of GARSN

Fig. 6 Training loss curves on PartNet dataset

Tab. 1 Comparison of cd loss for different algorithm on PartNet and Scan2CAD datasets

算法	PartNet				Scan2CAD
算法	chair	table	cabinet	平均	chair	table	cabinet	平均
Uy^［11］	2.02	2.32	2.63	2.22	3.36	6.65	7.26	4.90
ROCA^*［12］	2.50	2.72	3.86	2.72	4.24	14.97	15.92	9.15
ROCA^*［12］+De（U-RED）	3.80	3.87	2.82	3.76	6.99	8.10	13.08	8.10
ROCA^*［12］+De（VU-RED-F）	2.53	3.28	2.49	2.90	5.15	7.33	9.72	6.42
U-RED^［13］	0.95	1.33	1.30	1.17	2.89	3.16	5.95	3.35
PT43D^［3］	2.89	3.55	3.76	3.29	3.69	3.64	4.39	3.76
DiffCAD^［14］	—	—	—	—	6.40	7.20	6.90	6.71
VU-RED-F	0.81	1.10	1.03	0.97	2.43	2.62	4.96	2.80

Fig. 7 Training loss curves on ComplementMe dataset

Tab. 2 Comparison of cd loss on ComplementMe dataset

算法	chair	table	cabinet	平均
Uy^［11］	2.08	2.66	—	2.40
U-RED^［13］	1.68	2.26	—	2.00
VU-RED-F	1.37	1.91	—	1.67

Tab. 3 Computational efficiency comparison of different algorithms

算法	Params/10⁶	GFLOPs	Time/s
Uy^［11］	12.71	1.55	1.5
ROCA^*［12］	50.37	40.96	3.1
ROCA^*［12］+De（U-RED）	57.41	40.96	3.5
ROCA^*［12］+De（VU-RED-F）	57.93	40.96	3.5
U-RED^［13］	17.23	2.96	4.6
PT43D^［3］	75.97	$≥$ 100	9.7
DiffCAD^［14］	94.34	$≥$ 100	12.1
VU-RED-F	17.76	2.96	5.1

Tab. 3 Computational efficiency comparison of different algorithms

算法	Params/10⁶	GFLOPs	Time/s
Uy^［11］	12.71	1.55	1.5
ROCA^*［12］	50.37	40.96	3.1
ROCA^*［12］+De（U-RED）	57.41	40.96	3.5
ROCA^*［12］+De（VU-RED-F）	57.93	40.96	3.5
U-RED^［13］	17.23	2.96	4.6
PT43D^［3］	75.97	$≥$ 100	9.7
DiffCAD^［14］	94.34	$≥$ 100	12.1
VU-RED-F	17.76	2.96	5.1

Fig. 8 Visual comparison on PartNet dataset

Fig. 9 Visual comparison on Scan2CAD dataset

Tab. 4 Single-module test results on PartNet dataset

算法	chair	table	cabinet	平均
U-RED^［13］	0.95	1.33	1.30	1.17
VNE	0.88	1.19	1.18	1.06
FA-Res	0.91	1.27	1.15	1.11
GARSN	0.96	1.20	1.25	1.10
VU-RED-F	0.81	1.10	1.03	0.97

Tab. 5 Single-module test results on Scan2CAD dataset

算法	chair	table	cabinet	平均
U-RED^［13］	2.89	3.16	5.95	3.35
VNE	2.66	2.87	5.70	3.10
FA-Res	2.65	2.85	5.39	3.05
GARSN	2.73	2.91	5.62	3.14
VU-RED-F	2.43	2.62	4.96	2.80

Tab. 6 Ablation test results on PartNet dataset

算法	chair	table	cabinet	平均
U-RED^［13］	0.95	1.33	1.30	1.17
no VNE	0.85	1.13	1.09	1.01
no FA-Res	0.86	1.21	1.10	1.06
no GARSN	0.88	1.16	1.12	1.04
VU-RED-F	0.81	1.10	1.03	0.97

Tab. 7 Ablation test results on Scan2CAD dataset

算法	chair	table	cabinet	平均
U-RED^［13］	2.89	3.16	5.95	3.35
no VNE	2.55	2.79	5.24	2.96
no FA-Res	2.65	2.81	5.39	3.04
no GARSN	2.63	2.86	5.45	3.05
VU-RED-F	2.43	2.62	4.96	2.80

Tab. 8 Obstruction test results

遮挡比例/%	chair	table	cabinet	平均
0	0.75	1.07	0.99	0.93
25	0.78	1.08	1.02	0.95
50	0.81	1.10	1.03	0.97
75	0.93	1.54	1.20	1.26

References 32

[1]	邱云飞，赵静，方立. MF-Net：结合残差网络的多尺度特征点云补全网络［J］. 计算机工程与应用， 2023， 59（22）： 202-212.
	QIU Y F， ZHAO J， FANG L. MF-Net： multi-scale feature point cloud completion network combined with residual network［J］. Computer Engineering and Applications， 2023， 59（22）： 202-212.
[2]	MO S， XIE E， CHU R， et al. DiT-3D： exploring plain diffusion Transformers for 3D shape generation［C］// Proceedings of the 37th International Conference on Neural Information Processing Systems. Red Hook： Curran Associates Inc.， 2023： 67960-67971.
[3]	XIONG Y， DAI A. PT43D： a probabilistic Transformer for generating 3D shapes from single highly-ambiguous RGB images［C］// Proceedings of the 2024 British Machine Vision Conference. Durham： BMVA Press， 2024： No.77.
[4]	WU Y， YAN Z， CHEN C， et al. SCoDA： domain adaptive shape completion for real scans［C］// Proceedings of the 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2023： 17630-17641.
[5]	任世兴，韩燮，赵融，等. 一种由部件隐向量驱动的隐式三维重建方法［J］. 微电子学与计算机， 2025， 42（11）： 9-1.
	REN S X， HAN X， ZHAO R， et al. An implicit 3D reconstruction method driven by latent vectors of the components［J］. Microelectronics and Computer， 2025， 42（11）： 9-1.
[6]	RAHMAWATI D， SARNO R， FATICHAH C. Modification rules for improving marching cubes algorithm to represent 3D point cloud curve images［J］. International Journal of Intelligent Engineering and Systems， 2024， 17（1）： 536-554.
[7]	WU R， XIAO C， ZHENG C. DeepCAD： a deep generative network for computer-aided design models［C］// Proceedings of the 2021 IEEE/CVF International Conference on Computer Vision. Piscataway： IEEE， 2021： 6752-6762.
[8]	IRSHAD M Z， KOLLAR T， LASKEY M， et al. CenterSnap： single-shot multi-object 3D shape reconstruction and categorical 6D pose and size estimation［C］// Proceedings of the 2022 International Conference on Robotics and Automation. Piscataway： IEEE， 2022： 10632-10640.
[9]	SHUAI Q， ZHANG C， YANG K， et al. DPF-Net： combining explicit shape priors in deformable primitive field for unsupervised structural reconstruction of 3D objects［C］// Proceedings of the 2023 IEEE/CVF International Conference on Computer Vision. Piscataway： IEEE， 2023： 14275-14283.
[10]	UY M A， HUANG J， SUNG M， et al. Deformation-aware 3D model embedding and retrieval［C］// Proceedings of the 2020 European Conference on Computer Vision， LNCS 12352. Cham： Springer， 2020： 397-413.
[11]	UY M A， KIM V G， SUNG M， et al. Joint learning of 3D shape retrieval and deformation［C］// Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2021： 11708-11717.
[12]	GÜMELI C， DAI A， NIEßNER M. ROCA： robust CAD model retrieval and alignment from a single image［C］// Proceedings of the 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2022： 4012-4021.
[13]	DI Y， ZHANG C， ZHANG R， et al. U-RED： unsupervised 3D shape retrieval and deformation for partial point clouds［C］// Proceedings of the 2023 IEEE/CVF International Conference on Computer Vision. Piscataway： IEEE， 2023： 8850-8861.
[14]	GAO D， ROZENBERSZKI D， LEUTENEGGER S， et al. DiffCAD： weakly-supervised probabilistic CAD model retrieval and alignment from an RGB image［J］. ACM Transactions on Graphics， 2024， 43（4）： No.106.
[15]	QI C R， YI L， SU H， et al. PointNet++： deep hierarchical feature learning on point sets in a metric space［C］// Proceedings of the 31st International Conference on Neural Information Processing Systems. Red Hook： Curran Associates Inc.， 2017： 5105-5114.
[16]	WANG Y， SUN Y， LIU Z， et al. Dynamic graph CNN for learning on point clouds［J］. ACM Transactions on Graphics， 2019， 38（5）： No.146.
[17]	ZHAO H， JIANG L， JIA J， et al. Point Transformer［C］// Proceedings of the 2021 IEEE/CVF International Conference on Computer Vision. Piscataway： IEEE， 2021： 16239-16248.
[18]	GAO Y， LIU X， LI J， et al. LFT-Net： local feature Transformer network for point clouds analysis［J］. IEEE Transactions on Intelligent Transportation Systems， 2023， 24（2）： 2158-2168.
[19]	DENG C， LITANY O， DUAN Y， et al. Vector Neurons： a general framework for SO（3）-equivariant networks［C］// Proceedings of the 2021 IEEE/CVF International Conference on Computer Vision. Piscataway： IEEE， 2021： 12180-12189.
[20]	ZHANG Z， ZHANG L， WANG Z， et al. Part-level scene reconstruction affords robot interaction［C］// Proceedings of the 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems. Piscataway： IEEE， 2023： 11178-11185.
[21]	ZHAN L， LI W， MIN W. FA-ResNet： feature affine residual network for large-scale point cloud segmentation［J］. International Journal of Applied Earth Observation and Geoinformation， 2023， 118： No.103259.
[22]	HANOCKA R， HERTZ A， FISH N， et al. MeshCNN： a network with an edge［J］. ACM Transactions on Graphics， 2019， 38（4）： No.90.
[23]	THEKUMPARAMPIL K K， WANG C， OH S， et al. Attention-based graph neural network for semi-supervised learning［EB/OL］. ［2025-04-07］..
[24]	FAN H， SU H， GUIBAS L. A point set generation network for 3D object reconstruction from a single image［C］// Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2017： 2463-2471.
[25]	GOYAL A， BOCHKOVSKIY A， DENG J， et al. Non-deep networks［C］// Proceedings of the 36th International Conference on Neural Information Processing Systems. Red Hook： Curran Associates Inc.， 2021： 6789-6801.
[26]	BORAWAR L， KAUR R. ResNet： solving vanishing gradient in deep networks［C］// Proceedings of International Conference on Recent Trends in Computing： ICRTC 2022， LNNS 600. Singapore： Springer， 2023： 235-247.
[27]	YU S， PANG S， SONG L， et al. TSoft-Net： a novel transfer soft thresholding network based on self-attention for intelligent fault diagnosis of rotating machinery［J］. Measurement， 2024， 227： No.114237.
[28]	MO K， ZHU S， CHANG A X， et al. PartNet： a large-scale benchmark for fine-grained and hierarchical part-level 3D object understanding［C］// Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2019： 909-918.
[29]	CHANG A X， FUNKHOUSER T， GUIBAS L， et al. ShapeNet： an information-rich 3D model repository［EB/OL］. ［2025-04-07］..
[30]	SUNG M， SU H， KIM V G， et al. ComplementMe： weakly-supervised component suggestions for 3D modeling［J］. ACM Transactions on Graphics， 2017， 36（6）： No.226.
[31]	DAI A， CHANG A X， SAVVA M， et al. ScanNet： richly-annotated 3D reconstructions of indoor scenes［C］// Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2017： 2432-2443.
[32]	AVETISYAN A， DAHNERT M， DAI A， et al. Scan2CAD： learning CAD model alignment in RGB-D scans［C］// Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2019： 2609-2618.

VU-RED-F： improved CAD model replacement for U-RED single-view point clouds

VU-RED-F：改进U-RED的单视角点云CAD模型替换

RichHTML

PDF

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 17

References 32

Related Articles 15

Recommended Articles

Metrics

[1]	Wei ZHANG, Zhongwei GONG, Zhixin LI, Peihua LUO, Lingling SONG. Learning behavior boosted knowledge tracing model [J]. Journal of Computer Applications, 2025, 45(9): 2747-2754.
[2]	Zhiyu GONG, Shitong WANG. Learning method of residual network for heavy-tailed noisy image classification [J]. Journal of Computer Applications, 2025, 45(10): 3091-3100.
[3]	Yuanjiong LIU, Maozheng HE, Yibin HUANG, Cheng QIAN. Ship identification model based on ResNet50 and improved attention mechanism [J]. Journal of Computer Applications, 2024, 44(6): 1935-1941.
[4]	Lin GUO, Kunhu LIU, Chenyang MA, Youxue LAI, Yingfen XU. Image super-resolution reconstruction based on residual attention network with receptive field expansion [J]. Journal of Computer Applications, 2024, 44(5): 1579-1587.
[5]	Boyue WANG, Yingxiang LI, Jiandan ZHONG. Segmentation network for day and night ground-based cloud images based on improved Res-UNet [J]. Journal of Computer Applications, 2024, 44(4): 1310-1316.
[6]	Shunwang FU, Qian CHEN, Zhi LI, Guomei WANG, Yu LU. Two-channel progressive feature filtering network for tampered image detection and localization [J]. Journal of Computer Applications, 2024, 44(4): 1303-1309.
[7]	Jingxian ZHOU, Xina LI. UAV detection and recognition based on improved convolutional neural network and radio frequency fingerprint [J]. Journal of Computer Applications, 2024, 44(3): 876-882.
[8]	Xueyu HUANG, Huaiyu HE, Huimin LIN, Jinshui CHEN. Classification and recognition method of copper alloy metallograph based on feature aggregation [J]. Journal of Computer Applications, 2023, 43(8): 2593-2601.
[9]	Huibin ZHANG, Liping FENG, Yaojun HAO, Yining WANG. Ancient mural dynasty identification based on attention mechanism and transfer learning [J]. Journal of Computer Applications, 2023, 43(6): 1826-1832.
[10]	Lihua SHEN, Bo LI. Super-resolution reconstruction of lung CT images based on feature pyramid network and dense network [J]. Journal of Computer Applications, 2023, 43(5): 1612-1619.
[11]	Qihong SONG, Jianxun LIU, Haize HU, Xiangping ZHANG. Code search model based on collaborative fusion network [J]. Journal of Computer Applications, 2023, 43(12): 3896-3902.
[12]	Zhiang ZHANG, Guangzhong LIAO. Multi-scale feature enhanced retinal vessel segmentation algorithm based on U-Net [J]. Journal of Computer Applications, 2023, 43(10): 3275-3281.
[13]	Liefa LIAO, Zhiming LI, Saisai ZHANG. Image retrieval method based on deep residual network and iterative quantization hashing [J]. Journal of Computer Applications, 2022, 42(9): 2845-2852.
[14]	Huaiqing HE, Jianqing YAN, Kanghua HUI. Lightweight face recognition method based on deep residual network [J]. Journal of Computer Applications, 2022, 42(7): 2030-2036.
[15]	Yang ZHANG, Jiangbo HAO. Malicious code detection method based on attention mechanism and residual network [J]. Journal of Computer Applications, 2022, 42(6): 1708-1715.