L. Monaco

Fovea and Peripheral based Vision for Vision-Language Models.

AbstractIn this paper it is proposed a novel method that aims to achieve human-like vision (interms of efficiency) for VLMs. In doing so, an almost x10 token input reduction is achieved,and the training time is only 2-5 minutes on low end gpu hardware. Though, additionally,limita...

Fovea and Peripheral based Vision for Vision-Language Models.

AbstractIn this paper it is proposed a novel method that aims to achieve human-like vision (interms of efficiency) for VLMs. In doing so, an almost x10 token input reduction is achieved,and the training time is only 2-5 minutes on low end gpu hardware. Though, additionally,limita...