Researchers have introduced CogVis, a novel framework for open-vocabulary change detection (OVCD) in earth-surface monitoring. This new approach reformulates OVCD into a perception-memory-verification paradigm, decoupling temporal perception from semantic category decisions. CogVis utilizes a Scene Change Perceptron for change priors, a Semantic Memory Calibrator for score adjustments, and an Adaptive Region Filter for candidate refinement. The framework demonstrates state-of-the-art performance across seven benchmarks and improves inference throughput by 28.50% by sharing scene-level change perception. AI
IMPACT This research could lead to more efficient and accurate earth-surface monitoring systems by improving change detection capabilities.
RANK_REASON The cluster contains a research paper detailing a new AI framework and its performance on benchmarks. [lever_c_demoted from research: ic=1 ai=1.0]
- Adaptive Region Filter
- arXiv
- binary change localization
- building damage assessment
- Hugging Face
- Open-Vocabulary Change Detection
- Scene Change Perceptron
- semantic change detection
- Semantic Memory Calibrator
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
