Single Simple Patch
"Single simple patch" research explores the surprising effectiveness of small image regions in manipulating or analyzing complex visual data. Current work focuses on adversarial attacks using strategically placed patches to fool vision models (e.g., object detectors, image classifiers), as well as leveraging patches for improved image registration, anomaly detection, and even enhancing the efficiency of diffusion models. This research is significant because it reveals vulnerabilities in AI systems and offers novel approaches to improve model robustness, leading to more secure and reliable applications in diverse fields like autonomous driving and medical image analysis.
Papers
October 18, 2024
October 7, 2024
September 19, 2024
September 16, 2024
August 13, 2024
April 5, 2024
April 2, 2024
March 21, 2024
February 2, 2024
December 30, 2023
December 28, 2023
September 27, 2023
August 23, 2023
August 21, 2023
July 12, 2023
July 2, 2023
June 16, 2023
April 11, 2023
March 27, 2023