Robust Width
Robust width, a concept emerging across diverse fields, focuses on enhancing the resilience and accuracy of models, particularly in the face of noisy or adversarial inputs. Current research explores its application in improving the robustness of deep neural networks against adversarial attacks, as well as in image processing tasks like crack detection and centerline estimation in thick lines, often employing algorithms like Expectation-Maximization for parameter estimation. This research aims to improve model reliability and efficiency, impacting areas such as computer vision, medical imaging, and autonomous systems by providing more dependable and computationally efficient solutions.
Papers
May 24, 2024
April 3, 2024
September 10, 2022