Research map / Architectures and efficiency / CNN design
A Unified Model for Near and Remote Sensing
Scott Workman, Menghua Zhai, David J. Crandall, Nathan Jacobs
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 |
|---|---|---|---|
| fuse ground-level feature map with overhead CNN (unified uniform) instead of overhead only (remote) | remote · Brooklyn land use | top-1 accuracy: 69.63% -> 76.14% | Improved |
| fuse overhead features with ground-level feature map instead of ground-level only (proximate) | proximate · Brooklyn building age | top-1 accuracy: 35.90% -> 44.08% | Improved |
| use spatially varying adaptive kernel instead of uniform kernel (ablation) | unified · Brooklyn land use | top-1 accuracy: 76.14% -> 77.40% | Improved |
| use spatially varying adaptive kernel instead of uniform kernel (ablation) | unified · Queens land use | top-1 accuracy: 68.08% -> 70.55% | Improved |
| use adaptive kernel instead of uniform kernel (ablation) | unified · Brooklyn building age | top-1 accuracy: 44.08% -> 43.85% | No clear effect |
| use adaptive kernel instead of uniform kernel (ablation) | unified · Brooklyn building function | top-1 accuracy: 43.88% -> 44.88% | Improved |
| use adaptive kernel instead of uniform kernel (ablation) | unified · Queens building function | top-1 accuracy: 33.64% -> 34.13% | Improved |
| provide known (geo-oriented) ground image orientation instead of randomly circular-shifted orientation (ablation) | unified (adaptive) · Brooklyn land use | top-1 accuracy: 72.61% -> 77.40% | Improved |
| add ground-level images with unknown orientation to overhead-only model (ablation) | remote · Brooklyn land use | top-1 accuracy: about 3% higher than overhead only (72.61% vs 69.63%) | Improved |
| increase number of nearby ground images N up to 25 (ablation) | unified (adaptive) · Brooklyn land use | top-1 accuracy: saturated at N=15 | No clear effect |
| grid baseline (ground features on regular grid) vs unified adaptive | grid · Queens land use | top-1 accuracy: 67.51% -> 70.55% | Improved |
Closest papers by meaning
- Revisiting Near/Remote Sensing With Geospatial Attention (CVPR 2022 · similarity 0.88)
- Large Scale High-Resolution Land Cover Mapping With Multi-Resolution Data (CVPR 2019 · similarity 0.86)
- Geospecific View Generation - Geometry-Context Aware High-resolution Ground View Inference from Satellite Views (ECCV 2024 · similarity 0.86)
- Learning without Exact Guidance: Updating Large-scale High-resolution Land Cover Maps from Low-resolution Historical Labels (CVPR 2024 · similarity 0.85)
- SatSynth: Augmenting Image-Mask Pairs through Diffusion Models for Aerial Semantic Segmentation (CVPR 2024 · similarity 0.84)
- GeoViS: Geospatially Rewarded Visual Search for Remote Sensing Visual Grounding (CVPR 2026 · 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.
