Abstract: Large Vision-Language Models (LVLMs) suffer from severe object hallucinations, leading them to frequently generate outputs that do not correspond to the image content, significantly reducing ...
Abstract: Performing adversarial attacks on a visual tracker aims to drift the apparent target to the background by adding malicious perturbations to the source images. Demonstrating convincingly ...