Research map / Detection and segmentation / CNN features
Ultra-High Resolution Segmentation With Ultra-Rich Context: A Novel Benchmark
Deyi Ji, Feng Zhao, Hongtao Lu, Mingyuan Tao, Jieping Ye
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 |
|---|---|---|---|
| replace uniform bilinear downsampling with multi-level DWT-IWT in the deep branch (ablation) | WSDNet deep branch · URUR test | mIoU (%): +1.8 (45.1 to 46.9) | Improved |
| replace DWT downsampling with adaptive deformable-conv downsampling (ablation) | WSDNet deep branch · URUR test | mIoU (%): -0.9 (46.9 to 46.0), and FPS drops 7.13 to 4.96 | Hurt |
| add ordinary super-resolution loss instead of WSL (ablation) | WSDNet · URUR test | mIoU (%): +0.7 (45.2 to 45.9) | Improved |
| add Wavelet Smooth Loss (WSL) (ablation) | WSDNet · URUR test | mIoU (%): +1.7 (45.2 to 46.9) | Improved |
| WSDNet (evolution of ISDNet with DWT-IWT, WSL, RAF removed) vs ISDNet | segmentation on DeepGlobe · DeepGlobe test | mIoU (%): +0.8 (73.3 to 74.1) | Improved |
| WSDNet vs ISDNet | segmentation on Inria Aerial · Inria Aerial test | mIoU (%): +1.0 (74.2 to 75.2) | Improved |
| WSDNet vs ISDNet | segmentation on URUR · URUR test | mIoU (%): +1.1 (45.8 to 46.9) | Improved |
| removing RAF module (WSDNet vs ISDNet) raises inference speed | WSDNet · DeepGlobe test | FPS: 27.7 to 30.3 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ISDNet — ISDNet: Integrating Shallow and Deep Networks for Efficient Ultra-High Resolution Segmentation (CVPR 2022) · better on 4 of 4 shared results
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
- Ultra-High Resolution Segmentation via Boundary-Enhanced Patch-Merging Transformer (AAAI 2025 · similarity 0.87)
- UltraHR-100K: Enhancing UHR Image Synthesis with A Large-Scale High-Quality Dataset (NeurIPS 2025 · similarity 0.87)
- Boosting the Dual-Stream Architecture in Ultra-High Resolution Segmentation with Resolution-Biased Uncertainty Estimation (CVPR 2025 · similarity 0.87)
- UHR-BAT: Budget-Aware Token Compression Vision-Language model for Ultra-High-Resolution Remote Sensing (ICML 2026 · similarity 0.85)
- UltraLED: Learning to See Everything in Ultra-High Dynamic Range Scenes (NeurIPS 2025 · similarity 0.85)
- UltraVideo: High-Quality UHD Video Dataset with Comprehensive Captions (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.
