IMAGE OBJECT ANOMALY DETECTION

Ghulam Ahmed Ansari, Amrita Saha, Srikanth G. Tamilselvam

Abstract: This patent introduces an AI-driven system for automatic anomaly detection in drawings. The system analyzes user-generated sketches to identify factual inconsistencies based on commonsense knowledge (e.g., an ant being larger than a dog). It provides real-time feedback, highlights anomalies, and can automatically adjust the drawing for accuracy. The method leverages computer vision, deep learning, and generative models to ensure contextual correctness, making it useful for education, digital art critique, and automated content validation.