Research map / 3D vision / Face and shape
SeqXY2SeqZ: Structure Learning for 3D Shapes by Sequentially Predicting 1D Occupancy Segments From 2D Coordinates
Zhizhong Han, Guanhui Qiao, Yu-Shen Liu, Matthias Zwicker
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 |
|---|---|---|---|
| add attention (ablation) | SeqXY2SeqZ · ShapeNet chair single-image reconstruction | IoU: 47.5 -> 50.3 | Improved |
| use GRU cells instead of LSTM (ablation) | SeqXY2SeqZ · ShapeNet chair single-image reconstruction | IoU: 49.8 -> 50.3 | Improved |
| use bidirectional RNN encoder instead of single direction (ablation) | SeqXY2SeqZ · ShapeNet chair single-image reconstruction | IoU: 48.8 -> 50.3 | Improved |
| share one location embedding matrix across X, Y and Z (ablation) | SeqXY2SeqZ · ShapeNet chair single-image reconstruction | IoU: 50.3 -> 49.1 | No clear effect |
| tubelize along Z instead of Y (ablation) | SeqXY2SeqZ · ShapeNet airplane auto-encoding | IoU: 90.35 -> 89.96 | No clear effect |
| tubelize along X instead of Y (ablation) | SeqXY2SeqZ · ShapeNet airplane auto-encoding | IoU: 90.35 -> 90.21 | No clear effect |
| increase embedding and hidden size to 1024-2048 | SeqXY2SeqZ · ShapeNet chair auto-encoding | IoU: 74.32 -> 93.10 | Improved |
| use 2D-function tube prediction instead of 3D implicit function (DISN) | SeqXY2SeqZ · ShapeNet single-image 32^3 | mean IoU: 57.0 -> 63.6 | Improved |
| use 2D-function tube prediction instead of DIB-R | SeqXY2SeqZ · ShapeNet single-image 32^3 | mean IoU: 61.2 -> 63.6 | Improved |
| use 2D-function tube prediction instead of DISN | SeqXY2SeqZ · Single-image reconstruction cost | CPU time: 14.68s -> 8.79s | Improved |
| use 2D-function tube prediction instead of OccNet | SeqXY2SeqZ · Single-image reconstruction cost | memory: 1175 MB -> 286MB | Improved |
Closest papers by meaning
- Unsupervised Learning of Intrinsic Structural Representation Points (CVPR 2020 · similarity 0.88)
- Deep Functional Dictionaries: Learning Consistent Semantic Structures on 3D Models from Functions (NeurIPS 2018 · similarity 0.88)
- View-Aware Geometry-Structure Joint Learning for Single-View 3D Shape Reconstruction (TPAMI 2021 · similarity 0.88)
- RIM-Net: Recursive Implicit Fields for Unsupervised Learning of Hierarchical Shape Structures (CVPR 2022 · similarity 0.87)
- Matryoshka Networks: Predicting 3D Geometry via Nested Shape Layers (CVPR 2018 · similarity 0.87)
- Learning Partonomic 3D Reconstruction from Image Collections (CVPR 2025 · similarity 0.87)
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.
