Research map / Detection and segmentation / Open-set and few-shot detection
Semantic Understanding of Scenes Through the ADE20K Dataset
Bolei Zhou, Hang Zhao, Xavier Puig, Tete Xiao, Sanja Fidler, Adela Barriuso, Antonio B. Torralba
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 |
|---|---|---|---|
| use synchronized BN with batch/BN size 16 instead of frozen BN (ablation) | PSPNet DilatedResNet-50 · SceneParse150 val | mean IoU: 3.17% | Improved |
| reduce BN size from 4 to 2 (ablation) | PSPNet DilatedResNet-50 · SceneParse150 val | mean IoU: 0.4119 -> 0.3355 | Hurt |
| change batch size from 16 to 4 with synchronized BN (ablation) | PSPNet DilatedResNet-50 · SceneParse150 val | mean IoU: 0.4126 -> 0.4119 | No clear effect |
| increase batch size to 16 without synchronizing BN (BN size 2) (ablation) | PSPNet DilatedResNet-50 · SceneParse150 val | mean IoU: 0.3355 -> 0.3403 | No clear effect |
| vary batch size with frozen BN (ablation) | PSPNet DilatedResNet-50 · SceneParse150 val | mean IoU: 0.3809 -> 0.3856 | No clear effect |
| add cascade framework to SegNet | SegNet · SceneParse150 val | mean IoU: 0.2164 -> 0.2751 | Improved |
| add cascade framework to DilatedVGG | DilatedVGG · SceneParse150 val | mean IoU: 0.3231 -> 0.3490 | Improved |
| add Pyramid Pooling Module to DilatedResNet-50 (ablation) | DilatedResNet-50 · SceneParse150 val | mean IoU: 0.3549 -> 0.4204 | Improved |
| add Pyramid Pooling Module to DilatedResNet-18 (ablation) | DilatedResNet-18 · SceneParse150 val | mean IoU: 0.3534 -> 0.3800 | Improved |
| use multi-scale training for Mask R-CNN (ablation) | Mask R-CNN FPN-50 · InstSeg100 val | mAP: .1832 -> .2241 | Improved |
| fuse Mask R-CNN instance masks into scene parsing output | DilatedResNet-50 + PPM · SceneParse150 val | mean IoU: 0.4204 -> 0.4256 | Improved |
| fuse Mask R-CNN instance masks into scene parsing output | DilatedResNet-50 + PPM · SceneParse150 val | pixel accuracy: 80.23% -> 80.21% | No clear effect |
Closest papers by meaning
- Scene Parsing Through ADE20K Dataset (CVPR 2017 · similarity 0.93)
- Temporal Pixel-Level Semantic Understanding Through the VSPW Dataset (TPAMI 2023 · similarity 0.89)
- Hierarchy-Agnostic Unsupervised Segmentation: Parsing Semantic Image Structure (NeurIPS 2024 · similarity 0.87)
- SUM Parts: Benchmarking Part-Level Semantic Segmentation of Urban Meshes (CVPR 2025 · similarity 0.86)
- MaSS13K: A Matting-level Semantic Segmentation Benchmark (CVPR 2025 · similarity 0.86)
- OpenWorldSAM: Extending SAM2 for Universal Image Segmentation with Language Prompts (NeurIPS 2025 · 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.
