Researchers have developed CORE, a novel framework that leverages Large Multimodal Models (LMMs) for training-free anomaly synthesis in industrial settings. This approach addresses the scarcity of anomalous samples by using a prompt optimization technique that disentangles anomaly generation from preserving normal regions. CORE refines prompts through a self-supervised contrast-and-reflect mechanism, enabling LMMs to generate realistic anomalies for both zero-shot and one-shot scenarios without task-specific training. Experiments on MVTec-AD and VisA datasets show that CORE significantly improves downstream anomaly detection performance and can be integrated with various LMMs. AI
IMPACT Enhances industrial anomaly detection capabilities by enabling more realistic synthetic anomaly generation with existing LMMs.
RANK_REASON Academic paper detailing a new method for anomaly synthesis using LMMs. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
