Research map / Detection and segmentation / CNN features
AlignSeg: Feature-Aligned Segmentation Networks
Zilong Huang, Yunchao Wei, Xinggang Wang, Humphrey Shi, Wenyu Liu, Thomas S. Huang
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 Aligned Context Modeling (ablation) | bottom-up ResNet-101 baseline · Cityscapes val | mIoU (%): 77.1 -> 79.4 | Improved |
| add Aligned Feature Aggregation on top of AlignCM (ablation) | Baseline+AlignCM · Cityscapes val | mIoU (%): 79.4 -> 80.5 | Improved |
| add AlignFA alone (ablation) | bottom-up ResNet-101 baseline · Cityscapes val | mIoU (%): 77.1 -> 78.6 | Improved |
| replace regular grid sampling interpolation with AlignFA (ablation) | aggregation · Cityscapes val | mIoU (%): 79.4 -> 80.5 | Improved |
| replace deconvolution upsampling with AlignFA (ablation) | aggregation · Cityscapes val | mIoU (%): 79.7 -> 80.5 | Improved |
| align low-level feature with guide of high-level feature (ablation) | AlignFA · Cityscapes val | mIoU (%): 79.4 -> 79.7 | Improved |
| also align high-level feature (both directions) (ablation) | AlignFA · Cityscapes val | mIoU (%): 79.7 -> 80.5 | Improved |
| add alignment function to context module, bin size 2 (ablation) | AlignCM · Cityscapes val | mIoU (%): 79.1 -> 79.3 | Improved |
| add alignment function to context module, bin size 3 (ablation) | AlignCM · Cityscapes val | mIoU (%): 79.0 -> 79.4 | Improved |
| add alignment function to context module, bin size 6 (ablation) | AlignCM · Cityscapes val | mIoU (%): 78.5 -> 79.0 | Improved |
| AlignCM compared with attention-based RCCA (ablation) | context modeling · Cityscapes val | mIoU (%): 79.5 vs 79.4 | No clear effect |
| AlignCM compared with PSP and ASPP (ablation) | context modeling · Cityscapes val | mIoU (%): 79.0 and 79.1 -> 79.4 | Improved |
Later papers that compared with this method
2 later accepted papers report AlignSeg as a baseline in their results tables; 0 of them report a better number on most of the results they share with it.
- Learning Implicit Feature Alignment Function for Semantic Segmentation (ECCV 2022 · better on 0 of 1 shared result)
- AttaNet: Attention-Augmented Network for Fast and Accurate Scene Parsing (AAAI 2021 · better on 0 of 1 shared result)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- CCNet — CCNet: Criss-Cross Attention for Semantic Segmentation (ICCV 2019) · better on 2 of 2 shared results
- DANet — DANet: Divergent Activation for Weakly Supervised Object Localization (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
- AlignZeg: Mitigating Objective Misalignment for Zero-shot Semantic Segmentation (ECCV 2024 · similarity 0.88)
- Learning Implicit Feature Alignment Function for Semantic Segmentation (ECCV 2022 · similarity 0.87)
- PCA-Seg: Revisiting Cost Aggregation for Open-Vocabulary Semantic and Part Segmentation (CVPR 2026 · similarity 0.87)
- Improving Semi-Supervised Semantic Segmentation with Sliced-Wasserstein Feature Alignment and Uniformity (CVPR 2025 · similarity 0.86)
- Align and Segment: Unsupervised Learning for Building Segmentation From Misaligned Labels (ECCV 2026 · similarity 0.86)
- RankSEG-RMA: An Efficient Segmentation Algorithm via Reciprocal Moment Approximation (NeurIPS 2025 · 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.
