AD-Copilot: Comparison-Aware Anomaly Detection

AD-Copilot extends Qwen2.5-VL with a novel comparison-aware visual encoder that generates special comparison tokens capturing differences between a reference image and a test image, achieving state-of-the-art results on industrial anomaly detection benchmarks.

Two modes: Upload both images for comparison-based inspection, or just a test image for single-image tasks (counting, OCR, etc.).

[Paper] | [Code] | [Model]

16 1024
Examples
Reference (Good) Image (optional) Test Image Prompt Max New Tokens