limi124/remote-sensing-research-radar

Developer tools

128 stars Качество 70 Тренд 70

Repository Install

A Codex skill for tracking research frontiers in , , and .

Обзор

A Codex skill for tracking research frontiers in , , and .

README

Remote Sensing Research Radar

Language: English | 中文

A Codex skill for tracking research frontiers in geospatial AI, remote sensing big data, and transferable computer vision methods.

This project helps researchers continuously monitor new papers, open-source projects, datasets, benchmarks, and research trends. It focuses on optical remote sensing and geospatial AI, while also scanning computer vision methods that can be transferred into remote sensing research.

SAR, PolSAR, InSAR, radar-only, microwave-only, and SAR-optical fusion work are excluded by default unless explicitly requested.

Why This Skill

Remote sensing research moves quickly across multiple communities: geoscience, computer vision, machine learning, GIScience, and Earth observation. Important ideas often appear first in general CV or ML papers before being adapted to satellite and aerial imagery.

This skill is designed to bridge that gap:

Need	What the skill does
Track new papers	Searches recent arXiv, Papers with Code, GitHub, Hugging Face, conference pages, and benchmark pages
Find useful CV methods	Screens computer vision papers for remote sensing transfer potential
Avoid noisy directions	Filters out SAR-focused work by default
Build a reading queue	Ranks papers and projects by novelty, evidence, reproducibility, and fit
Generate research ideas	Converts frontier trends into publishable research directions
Prepare related work	Produces comparison tables and structured literature summaries

Core Capabilities

Daily or weekly research radar reports
Paper and GitHub project screening
Remote sensing frontier tracking
Computer-vision-to-remote-sensing transfer analysis
Related work table generation
Deep paper reading notes
Reproducibility planning
Publishable idea generation

Research Scope

Remote Sensing and Geospatial AI

Area	Example Topics
Earth observation foundation models	MAE, JEPA, contrastive learning, masked modeling, geolocation/time embeddings
Optical and multispectral imagery	Sentinel-2, Landsat, aerial RGB, UAV imagery, VHR scene understanding
Hyperspectral learning	Spectral-spatial modeling, band selection, domain generalization
Multi-temporal modeling	Crop monitoring, land-cover dynamics, urban expansion, ecological time series
Change detection	Optical building, land-cover, urban, and ecological change detection
Remote sensing VLMs	Captioning, VQA, grounding, open-vocabulary segmentation
Cross-domain generalization	Domain adaptation, OOD detection, few-shot learning, spatial bias
GIS and raster-vector fusion	Road networks, parcels, POIs, building footprints, topology-aware learning
Efficient inference	Tiling, streaming, compression, active learning, uncertainty-aware mapping

Transferable Computer Vision

CV Direction	Remote Sensing Transfer Path
Foundation models	Adapt pretrained visual representations to multispectral, high-resolution, or multi-temporal imagery
Open-vocabulary segmentation	Map flexible land-cover classes and local taxonomy labels
Vision-language models	Support remote sensing VQA, captioning, grounding, and retrieval
Video and temporal models	Transfer long-range temporal reasoning to satellite image time series
Domain adaptation	Improve cross-city, cross-country, cross-season, and cross-sensor robustness
Weak/semi-supervised learning	Reduce annotation cost for large-scale mapping
Efficient inference	Handle large remote sensing images with tiled or compressed inference

Scoring Rubric

Candidate papers and projects are ranked by:

Criterion	Meaning
Novelty	New problem, method, benchmark, or strong reframing
Technical depth	Clear algorithmic or modeling contribution
Evidence	Strong experiments, ablations, datasets, and metrics
Reproducibility	Code, data, weights, and clean instructions
Trend signal	Active authors, labs, stars, benchmarks, or conference attention
Transferability	Clear path from CV or ML to remote sensing constraints
User fit	Matches non-SAR geospatial remote sensing research goals

Installation

Install the skill from this repository:

python C:\Users\shuo\.codex\skills\.system\skill-installer\scripts\install-skill-from-github.py --repo limi124/remote-sensing-research-radar --path skills/remote-sensing-research-radar

Restart Codex after installation so the skill can be discovered.

Recommended Repository Structure

remote-sensing-research-radar/
├─ README.md
├─ README.zh-CN.md
└─ skills/
   └─ remote-sensing-research-radar/
      ├─ SKILL.md
      ├─ agents/
      │  └─ openai.yaml
      └─ references/
         └─ frontier-radar.md

Example Prompts

Use remote-sensing-research-radar to find recent remote sensing foundation model papers and open-source projects from the past month. Exclude SAR.

Build a weekly geospatial AI research radar. Focus on optical remote sensing, multi-temporal modeling, and transferable computer vision methods.

Find computer vision papers on open-vocabulary segmentation, VLMs, and domain adaptation that could inspire remote sensing research.

Based on recent remote sensing and CV frontiers, propose five publishable research ideas for a high-impact paper.

Typical Output

Weekly takeaways
- Trend 1
- Trend 2
- Trend 3

Ranked candidates
| Rank | Title | Source/Date | Task | Data/Modality | Contribution | Code/Data | Score | Why it matters |

Top 3 deep dives
1. Title
   - Core problem:
   - Method:
   - Evidence:
   - Limitations:
   - Extension:

CV-to-RS transfer notes
- Paper/project:
- Transferable component:
- Required adaptation:
- Candidate RS datasets:
- First experiment:
- Risk:

Publishable research ideas
1. Idea name
   - Problem:
   - Hypothesis:
   - Method:
   - Datasets/Metrics:
   - Baselines:
   - First experiment:

Notes

Fresh research reports should use web search because arXiv, GitHub, Papers with Code, and conference pages change frequently.
English paper titles and technical terms should be preserved even in Chinese reports.
SAR-related work is excluded by default to keep the radar aligned with optical remote sensing and geospatial AI research.

View this README on GitHub

Установка

npx skillfish add limi124/remote-sensing-research-radar