Research map / Large language models / Alignment and preferences
Human-Machine Collaboration for Fast Land Cover Mapping
Read the paperCheck 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 human-guided point selection and labeling instead of random point selection | Maryland U-net fine-tuning with Last 2 Layers · New York target areas, within 3 minutes (approximately 60 samples) | accuracy | Helped |
| Use human-guided point selection and labeling instead of random point selection | Maryland U-net fine-tuning with Last 2 Layers · New York target areas, within 3 minutes (approximately 60 samples) | mean IoU | Helped |
| Use Min-Margin instead of Random to select labeled points | Maryland U-net fine-tuning methods · New York target areas, 400 labeled pixels | IoU: 0.608 -> 0.608; 0.677 -> 0.701; 0.658 -> 0.685; 0.562 -> 0.580; 0.597 -> 0.601 | Mixed |
| Use Entropy instead of Random to select labeled points | Maryland U-net fine-tuning methods · New York target areas, 400 labeled pixels | IoU: 0.608 -> 0.501; 0.677 -> 0.587; 0.658 -> 0.572; 0.562 -> 0.520; 0.597 -> 0.550 | Hurt |
| Select points where the model makes errors instead of random points | Maryland U-net fine-tuning methods · New York target areas, 400 labeled pixels | IoU: 0.608 -> 0.551; 0.677 -> 0.631; 0.658 -> 0.621; 0.562 -> 0.575; 0.597 -> 0.609 | Mixed |
| Fine-tune the last two layers rather than the last one | Maryland U-net with Random query selection · New York target areas, 400 labeled pixels | IoU: 0.608 -> 0.677 | Helped |
| Fine-tune the last three layers rather than the last two | Maryland U-net with Random query selection · New York target areas, 400 labeled pixels | IoU: 0.677 -> 0.658 | Hurt |
| Fine-tune group normalization parameters rather than the last two layers | Maryland U-net with Random query selection · New York target areas, 400 labeled pixels | IoU: 0.677 -> 0.562 | Hurt |
| Use dropout-mask optimization rather than fine-tuning the last two layers | Maryland U-net with Random query selection · New York target areas, 400 labeled pixels | IoU: 0.677 -> 0.597 | Hurt |
| Spend more time using the interactive tool to select and label points | Human labelers adapting the Maryland U-net · New York target areas | model performance | Helped |
Closest papers by meaning
- Learning without Exact Guidance: Updating Large-scale High-resolution Land Cover Maps from Low-resolution Historical Labels (CVPR 2024 · similarity 0.87)
- DUNIA: Pixel-Sized Embeddings via Cross-Modal Alignment for Earth Observation Applications (ICML 2025 · similarity 0.83)
- Fields of The World: A Machine Learning Benchmark Dataset for Global Agricultural Field Boundary Segmentation (AAAI 2025 · similarity 0.83)
- InstructSAM: A Training-free Framework for Instruction-Oriented Remote Sensing Object Recognition (NeurIPS 2025 · similarity 0.83)
- SatSynth: Augmenting Image-Mask Pairs through Diffusion Models for Aerial Semantic Segmentation (CVPR 2024 · similarity 0.82)
- ZoomEarth: Active Perception for Ultra-High-Resolution Geospatial Vision-Language Tasks (CVPR 2026 · similarity 0.82)
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.