Research map / Detection and segmentation / CNN features
SPGNet: Semantic Prediction Guidance for Scene Parsing
Bowen Cheng, Liang-Chieh Chen, Yunchao Wei, Yukun Zhu, Zilong Huang, Jinjun Xiong, Thomas S. Huang, Wen-Mei Hwu, Honghui Shi
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 a second encoder-decoder stage (ablation) | 1-stage encoder-decoder · Cityscapes val (ResNet-18) | mIoU: 74.48 -> 76.31 | Improved |
| add SPG module with sigmoid supervise-and-excite (ablation) | 2-stage network · Cityscapes val (ResNet-18) | mIoU: 76.31 -> 77.67 | Improved |
| use sigmoid instead of softmax excitation (ablation) | SPG · Cityscapes val (ResNet-18) | mIoU: 77.17 -> 77.67 | Improved |
| use excitation instead of simply summing predictions (ablation) | SPG · Cityscapes val (ResNet-18) | mIoU: 76.96 -> 77.67 | Improved |
| add identity mapping path (ablation) | SPG · Cityscapes val (ResNet-18) | mIoU: 77.24 -> 77.67 | Improved |
| add semantic supervision to guided attention (ablation) | SPG · Cityscapes val (ResNet-18) | mIoU: 77.12 -> 77.67 | Improved |
| replace unsupervised GE module with supervised SPG (ablation) | 2-stage network · Cityscapes val (ResNet-18) | mIoU: 77.22 -> 77.67 | Improved |
| add a third stage (ablation) | SPGNet · Cityscapes val (ResNet-18) | mIoU: 77.67 -> 77.66 | No clear effect |
| use deeper backbone in the first encoder (ablation) | 2-stage SPGNet · Cityscapes val | mIoU: 78.34 -> 78.83 | Improved |
| add OHEM (ablation) | 2-stage SPGNet R101 · Cityscapes val | mIoU: 80.04 -> 80.85 | Improved |
| increase decoder channels from 128 to 256 (ablation) | 2-stage SPGNet · Cityscapes val | mIoU: 80.10 -> 80.91 (R50), 80.85 -> 80.42 (R101) | Mixed |
| add multi-scale and flip testing (ablation) | SPGNet R50 · Cityscapes val | mIoU: 80.91 -> 81.86 | Improved |
Later papers that compared with this method
1 later accepted paper reports SPGNet as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- SemIE: Semantically-Aware Image Extrapolation (ICCV 2021 · better on 13 of 14 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 3 on most shared results.
- DANet — DANet: Divergent Activation for Weakly Supervised Object Localization (ICCV 2019) · better on 0 of 1 shared result
- BiSeNet — BiSeNet: Bilateral Segmentation Network for Real-time Semantic Segmentation (ECCV 2018) · better on 1 of 1 shared result
- PSANet — PSANet: Point-wise Spatial Attention Network for Scene Parsing (ECCV 2018) · 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
- SegNet: A Deep Convolutional Encoder-Decoder Architecture for Image Segmentation (TPAMI 2017 · similarity 0.89)
- SegNeXt: Rethinking Convolutional Attention Design for Semantic Segmentation (NeurIPS 2022 · similarity 0.88)
- Rethinking Semantic Segmentation From a Sequence-to-Sequence Perspective With Transformers (CVPR 2021 · similarity 0.87)
- SegMAN: Omni-scale Context Modeling with State Space Models and Local Attention for Semantic Segmentation (CVPR 2025 · similarity 0.86)
- SGDiff: Scene Graph Guided Diffusion Model for Image Collaborative SegCaptioning (AAAI 2025 · similarity 0.84)
- OmniSegmentor: A Flexible Multi-Modal Learning Framework for Semantic Segmentation (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.
