Lightweight Explicit 3D Human Digitization via Normal Integration
In recent years, generating 3D human models from images has gained significant attention in 3D human reconstruction. However, deploying large neural network models in practical applications remains challenging, particularly on resource-constrained edge devices. This problem is primarily because larg...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-02-01
|
| Series: | Sensors |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1424-8220/25/5/1513 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1850222724908056576 |
|---|---|
| author | Jiaxuan Liu Jingyi Wu Ruiyang Jing Han Yu Jing Liu Liang Song |
| author_facet | Jiaxuan Liu Jingyi Wu Ruiyang Jing Han Yu Jing Liu Liang Song |
| author_sort | Jiaxuan Liu |
| collection | DOAJ |
| description | In recent years, generating 3D human models from images has gained significant attention in 3D human reconstruction. However, deploying large neural network models in practical applications remains challenging, particularly on resource-constrained edge devices. This problem is primarily because large neural network models require significantly higher computational power, which imposes greater demands on hardware capabilities and inference time. To address this issue, we can optimize the network architecture to reduce the number of model parameters, thereby alleviating the heavy reliance on hardware resources. We propose a lightweight and efficient 3D human reconstruction model that balances reconstruction accuracy and computational cost. Specifically, our model integrates Dilated Convolutions and the Cross-Covariance Attention mechanism into its architecture to construct a lightweight generative network. This design effectively captures multi-scale information while significantly reducing model complexity. Additionally, we introduce an innovative loss function tailored to the geometric properties of normal maps. This loss function provides a more accurate measure of surface reconstruction quality and enhances the overall reconstruction performance. Experimental results show that, compared with existing methods, our approach reduces the number of training parameters by approximately 80% while maintaining the generated model’s quality. |
| format | Article |
| id | doaj-art-702fb016dfd643c689cb14b25538db5f |
| institution | OA Journals |
| issn | 1424-8220 |
| language | English |
| publishDate | 2025-02-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Sensors |
| spelling | doaj-art-702fb016dfd643c689cb14b25538db5f2025-08-20T02:06:15ZengMDPI AGSensors1424-82202025-02-01255151310.3390/s25051513Lightweight Explicit 3D Human Digitization via Normal IntegrationJiaxuan Liu0Jingyi Wu1Ruiyang Jing2Han Yu3Jing Liu4Liang Song5Academy for Engineering and Technology, Fudan University, Shanghai 200433, ChinaAcademy for Engineering and Technology, Fudan University, Shanghai 200433, ChinaAcademy for Engineering and Technology, Fudan University, Shanghai 200433, ChinaAcademy for Engineering and Technology, Fudan University, Shanghai 200433, ChinaAcademy for Engineering and Technology, Fudan University, Shanghai 200433, ChinaAcademy for Engineering and Technology, Fudan University, Shanghai 200433, ChinaIn recent years, generating 3D human models from images has gained significant attention in 3D human reconstruction. However, deploying large neural network models in practical applications remains challenging, particularly on resource-constrained edge devices. This problem is primarily because large neural network models require significantly higher computational power, which imposes greater demands on hardware capabilities and inference time. To address this issue, we can optimize the network architecture to reduce the number of model parameters, thereby alleviating the heavy reliance on hardware resources. We propose a lightweight and efficient 3D human reconstruction model that balances reconstruction accuracy and computational cost. Specifically, our model integrates Dilated Convolutions and the Cross-Covariance Attention mechanism into its architecture to construct a lightweight generative network. This design effectively captures multi-scale information while significantly reducing model complexity. Additionally, we introduce an innovative loss function tailored to the geometric properties of normal maps. This loss function provides a more accurate measure of surface reconstruction quality and enhances the overall reconstruction performance. Experimental results show that, compared with existing methods, our approach reduces the number of training parameters by approximately 80% while maintaining the generated model’s quality.https://www.mdpi.com/1424-8220/25/5/1513three-dimensional human reconstructionnormal map estimationa skinned multi-person linear modeldeep learning |
| spellingShingle | Jiaxuan Liu Jingyi Wu Ruiyang Jing Han Yu Jing Liu Liang Song Lightweight Explicit 3D Human Digitization via Normal Integration Sensors three-dimensional human reconstruction normal map estimation a skinned multi-person linear model deep learning |
| title | Lightweight Explicit 3D Human Digitization via Normal Integration |
| title_full | Lightweight Explicit 3D Human Digitization via Normal Integration |
| title_fullStr | Lightweight Explicit 3D Human Digitization via Normal Integration |
| title_full_unstemmed | Lightweight Explicit 3D Human Digitization via Normal Integration |
| title_short | Lightweight Explicit 3D Human Digitization via Normal Integration |
| title_sort | lightweight explicit 3d human digitization via normal integration |
| topic | three-dimensional human reconstruction normal map estimation a skinned multi-person linear model deep learning |
| url | https://www.mdpi.com/1424-8220/25/5/1513 |
| work_keys_str_mv | AT jiaxuanliu lightweightexplicit3dhumandigitizationvianormalintegration AT jingyiwu lightweightexplicit3dhumandigitizationvianormalintegration AT ruiyangjing lightweightexplicit3dhumandigitizationvianormalintegration AT hanyu lightweightexplicit3dhumandigitizationvianormalintegration AT jingliu lightweightexplicit3dhumandigitizationvianormalintegration AT liangsong lightweightexplicit3dhumandigitizationvianormalintegration |