OneFocus: Enabling Real-World X-ray Security Screening with a Unified Vision-Language Model
Researchers have introduced OneFocus, a novel vision-language model designed to enhance X-ray security screening capabilities. To address the scarcity of relevant training data, they developed MMXray, a benchmark dataset comprising over 52,000 image-caption pairs of X-ray contraband, alongside CleanDET and AnyContraSyn for synthetic data generation. OneFocus is built to perform multiple tasks including visual question answering, contraband localization, and classification, aiming to improve generalization and understanding in security applications. AI
IMPACT This research could lead to more effective automated contraband detection systems, improving security in logistics and transportation.