The human brain extracts complex information from visual inputs, including objects, their spatial and semantic interrelations, and their interactions with the environment. However, a quantitative ...

Abstract: Most state-of-the-art trackers use variants of the Vision Transformer (ViT) as backbone. Nevertheless, their applicability is often constrained by large model sizes. Vision Mamba (Vim) ...

Change: We will modify the reward for touching the ball and the existential reward/penalty to increase as the game progresses. This makes the agent more aggressive and strategic later in the episode.