Research map / Detection and segmentation / CNN features
Tensor Low-Rank Reconstruction for Semantic Segmentation
Wanli Chen, Xinge Zhu, Ruoqi Sun, Junjun He, Ruiyu Li, Xiaoyong Shen, Bei Yu
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 TGM+TRM tensor low-rank reconstruction (ablation) | ResNet-50 baseline · PASCAL VOC12 val | mIoU (%): 68.7 -> 78.6 | Improved |
| add global pooling module (GPM) (ablation) | ResNet-50 + TGM+TRM · PASCAL VOC12 val | mIoU (%): 78.6 -> 79.2 | Improved |
| add auxiliary loss after Res-4 (ablation) | ResNet-50 + TGM+TRM + GPM · PASCAL VOC12 val | mIoU (%): 79.2 -> 79.8 | Improved |
| use ResNet-101 instead of ResNet-50 (ablation) | RecoNet · PASCAL VOC12 val | mIoU (%): 79.8 -> 81.4 | Improved |
| add multi-scale and flip testing (ablation) | RecoNet ResNet-101 · PASCAL VOC12 val | mIoU (%): 81.4 -> 82.1 | Improved |
| fine-tune on VOC12 original train set (ablation) | RecoNet ResNet-101 · PASCAL VOC12 val | mIoU (%): 82.1 -> 82.9 | Improved |
| increase tensor rank from 16 to 64 (ablation) | RecoNet · PASCAL VOC12 val | mIoU (%): 81.2 -> 82.1 | Improved |
| increase tensor rank from 64 to 128 (ablation) | RecoNet · PASCAL VOC12 val | mIoU (%): 82.1 -> 80.7 | Hurt |
| add TGM+TRM (ablation) | ResNet-50 baseline (Cityscapes val) · Cityscapes val | mIoU (%): 73.1 -> 78.9 | Improved |
| use RecoNet context module instead of EMANet (ablation) | ResNet-101 (single scale) · PASCAL VOC12 val | mIoU (%): 80.09 -> 81.40 | Improved |
| use RecoNet instead of DANet context module | ResNet-101 (single scale) · PASCAL VOC12 val | mIoU (%): 79.64 -> 81.40 | Improved |
| use TGM+TRM instead of non-local block | context module cost · computational cost (512 channels) | FLOPs: 19.33G -> 0.0215G | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- CCNet — CCNet: Criss-Cross Attention for Semantic Segmentation (ICCV 2019) · better on 1 of 1 shared result
- DANet — DANet: Divergent Activation for Weakly Supervised Object Localization (ICCV 2019) · better on 1 of 1 shared result
- ACFNet — ACFNet: Attentional Class Feature Network for Semantic Segmentation (ICCV 2019) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Rewis3d: Reconstruction Improves Weakly-Supervised Semantic Segmentation (CVPR 2026 · similarity 0.86)
- Performance Gaps in Multi-view Clustering under the Nested Matrix-Tensor Model (ICLR 2024 · similarity 0.85)
- Hierarchy-Agnostic Unsupervised Segmentation: Parsing Semantic Image Structure (NeurIPS 2024 · similarity 0.84)
- LowRankOcc: Tensor Decomposition and Low-Rank Recovery for Vision-based 3D Semantic Occupancy Prediction (CVPR 2024 · similarity 0.84)
- SegMAN: Omni-scale Context Modeling with State Space Models and Local Attention for Semantic Segmentation (CVPR 2025 · similarity 0.84)
- Rethinking Decoders for Transformer-based Semantic Segmentation: A Compression Perspective (NeurIPS 2024 · similarity 0.84)
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.
