Research map / Architectures and efficiency / CNN design
T-Net: Parametrizing Fully Convolutional Nets With a Single High-Order Tensor
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Parametrize the whole network with one low-rank Tucker tensor instead of leaving it uncompressed | Stacked HourGlass · MPII validation | PCKh: 87% -> 87.5% | Helped |
| Use a single network-level tensor rather than layer-wise decomposition | Stacked HourGlass · MPII validation | PCKh: 85.3% -> 87.1% | Helped |
| Use network-level tensorization rather than trimming channels | Stacked HourGlass · MPII validation | PCKh: 84.5% -> 87.1% | Helped |
| Reduce the #hg rank to 3 (ablation) | Tucker T-Net · MPII validation | PCKh: 86.99% -> 87.42% | Helped |
| Reduce the #hg rank to 1 (ablation) | Tucker T-Net · MPII validation | PCKh: 86.99% -> 86.05% | Hurt |
| Reduce hg_depth rank to 3 (ablation) | Tucker T-Net · MPII validation | PCKh: 86.99% -> 87.71% | Helped |
| Reduce hg_subnet rank to 1 (ablation) | Tucker T-Net · MPII validation | PCKh: 86.99% -> 86.19% | Hurt |
| Reduce b_depth rank to 1 (ablation) | Tucker T-Net · MPII validation | PCKh: 86.99% -> 82.59% | Hurt |
| Reduce feature ranks to 96 (ablation) | Tucker T-Net · MPII validation | PCKh: 86.99% -> 87.43% | Helped |
| Reduce feature ranks to 32 (ablation) | Tucker T-Net · MPII validation | PCKh: 86.99% -> 83.10% | Hurt |
| Reduce kernel height and width ranks to 2 (ablation) | Tucker T-Net · MPII validation | PCKh: 86.99% -> 87.30% | Helped |
| Apply Tucker tensorization at 3.38x compression | Fully Convolutional Network · Facial part segmentation test | mIOU: 76.02% -> 76.01% | No clear effect |
Closest papers by meaning
- Convolutions and More as Einsum: A Tensor Network Perspective with Advances for Second-Order Methods (NeurIPS 2024 · similarity 0.89)
- Distribution-Aware Tensor Decomposition for Compression of Convolutional Neural Networks (NeurIPS 2025 · similarity 0.88)
- Geometry-aware training of factorized layers in tensor Tucker format (NeurIPS 2024 · similarity 0.87)
- Tensor Programs VI: Feature Learning in Infinite Depth Neural Networks (ICLR 2024 · similarity 0.87)
- Coarse-To-Fine Tensor Trains for Compact Visual Representations (ICML 2024 · similarity 0.86)
- Over-parameterized Student Model via Tensor Decomposition Boosted Knowledge Distillation (NeurIPS 2024 · similarity 0.86)
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.