A Fast Geometric Regularizer to Mitigate Event Collapse in the Contrast Maximization Framework

Event cameras are emerging vision sensors and their advantages suitable for various applications such as autonomous robots. Contrast maximization (CMax), which provides state‐of‐the‐art accuracy on motion estimation using events, may suffer from an overfitting problem called event collapse. Prior works computationally expensive or cannot alleviate the overfitting, undermines benefits of CMax framework. A novel, efficient regularizer based geometric principles to mitigate collapse is proposed. The experiments show that proposed achieves results, while its reduced computational complexity makes it two four times faster than previous approaches. To best our knowledge, this only effective solution without trading off runtime. It hoped work opens door future unlocks cameras. Project page: https://github.com/tub‐rip/event_collapse