Research map / 3D vision / Depth and stereo
Deep Volumetric Video From Very Sparse Multi-View Performance Capture
Zeng Huang, Tianye Li, Weikai Chen, Yajie Zhao, Jun Xing, Chloe LeGendre, Linjie Luo, Chongyang Ma, Hao Li
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| increase number of input views from 3 to 8 (train and test per count) (ablation) | multi-view volumetric CNN · View-count study, left subject | median mesh-to-scan distance (cm): 0.84cm (three-view) -> 0.45cm (eight-view) | Improved |
| increase number of input views from 3 to 8 (train and test per count) (ablation) | multi-view volumetric CNN · View-count study, right subject | median mesh-to-scan distance (cm): 1.38cm (three-view) -> 0.59cm (eight-view) | Improved |
| increase views from 3 to 4 (ablation) | multi-view volumetric CNN · View-count study, right subject | median mesh-to-scan distance (cm): 1.38cm -> 1.06cm | Improved |
| use ours instead of PMVS/SurfaceNet/visual hull with four views (qualitative) | 4-view reconstruction · Four-view comparison (Fig 7) | reconstruction completeness and shape fidelity (qualitative) | Improved |
| train with a small set of random backgrounds instead of green screen only (ablation) | multi-view volumetric CNN · complex-background inputs (Fig 8) | reconstruction quality (qualitative) | Hurt |
| add the octree coarse-to-fine reconstruction, 256^3 vs 512^3 resolution (ablation) | inference · 4-view runtime | runtime (s): 4.4 seconds -> 18 seconds | Hurt |
Closest papers by meaning
- Volumetric performance capture from minimal camera viewpoints (ECCV 2018 · similarity 0.90)
- NeuralHumanFVV: Real-Time Neural Volumetric Human Performance Rendering Using RGB Cameras (CVPR 2021 · similarity 0.89)
- Volumetric Capture of Humans With a Single RGBD Camera via Semi-Parametric Learning (CVPR 2019 · similarity 0.89)
- A Deeper Look into DeepCap (TPAMI 2021 · similarity 0.88)
- Shape Reconstruction Using Volume Sweeping and Learned Photoconsistency (ECCV 2018 · similarity 0.88)
- DeepMultiCap: Performance Capture of Multiple Characters Using Sparse Multiview Cameras (ICCV 2021 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
