Research map / Detection and segmentation / Open-set and few-shot detection
The Cityscapes Dataset for Semantic Urban Scene Understanding
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 |
|---|---|---|---|
| Use full-resolution rather than half-resolution input (ablation) | FCN-8s · Cityscapes test | IoU class: 61.9 -> 65.3 | Helped |
| Use full-resolution rather than half-resolution input (ablation) | FCN-8s · Cityscapes test | iIoU class: 33.6 -> 41.7 | Helped |
| Train with train plus validation fine annotations rather than validation annotations only (ablation) | FCN-8s · Cityscapes test | IoU class: 58.3 -> 65.3 | Helped |
| Use fine annotations rather than coarse annotations only (ablation) | FCN-8s · Cityscapes test | IoU class: 58.0 -> 65.3 | Helped |
| Use fine annotations rather than coarse annotations only (ablation) | FCN-8s · Cityscapes test | iIoU class: 31.8 -> 41.7 | Helped |
| Use FCN-8s rather than FCN-16s (ablation) | FCN semantic labeling · Cityscapes test | IoU class: 64.3 -> 65.3 | Helped |
| Use FCN-16s rather than FCN-32s (ablation) | FCN semantic labeling · Cityscapes test | IoU class: 61.3 -> 64.3 | Helped |
| Use coarse annotations as additional training data with an appropriate method (ablation) | DeepLab · Cityscapes test | IoU class: 64.8 -> 66.4 | Helped |
| Replace MCG proposals with ground-truth regions (ablation) | Fast R-CNN instance labeling · Cityscapes instance-level labeling test | AP: 2.6 -> 41.3 | Helped |
| Replace Fast R-CNN classifier with an oracle while keeping MCG regions (ablation) | Fast R-CNN instance labeling · Cityscapes instance-level labeling test | AP: 2.6 -> 10.5 | Helped |
| Use MCG convex hulls rather than MCG regions (ablation) | Fast R-CNN instance labeling · Cityscapes instance-level labeling test | AP: 2.6 -> 4.6 | Helped |
| Train on Cityscapes and evaluate without fine-tuning on CamVid | FCN-8s · CamVid | average recall: 62.9 -> 72.6 | Helped |
Closest papers by meaning
- UrbanFeel：A Comprehensive Benchmark for Temporal and Perceptual Understanding of City Scenes through Human Perspective (ICLR 2026 · similarity 0.87)
- SUM Parts: Benchmarking Part-Level Semantic Segmentation of Urban Meshes (CVPR 2025 · similarity 0.87)
- Weakly Supervised Semantic Segmentation for Driving Scenes (AAAI 2024 · similarity 0.87)
- Improving Region Representation Learning from Urban Imagery with Noisy Long-Caption Supervision (AAAI 2026 · similarity 0.86)
- Learning Content-Enhanced Mask Transformer for Domain Generalized Urban-Scene Segmentation (AAAI 2024 · similarity 0.86)
- 3D StreetUnveiler with Semantic-aware 2DGS - a simple baseline (ICLR 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.