The field of autonomous perception is witnessing significant advancements, driven by the development of efficient and accurate object detection and tracking systems. Researchers are exploring innovative approaches to optimize performance in computationally constrained environments, such as autonomous racing and surveillance. Notably, there is a growing focus on enhancing temporal modeling capabilities and fusion strategies to improve 3D detection and semantic occupancy prediction. These advancements have the potential to revolutionize various applications, including autonomous driving, traffic monitoring, and urban surveillance. Noteworthy papers in this area include: TinyCenterSpeed, which introduces a streamlined adaptation of the CenterPoint method for real-time performance on autonomous racing platforms. PatrolVision, which presents a novel prototype for automated license plate recognition in urban environments. RoPETR, which proposes a customized positional embedding strategy to enhance temporal modeling capabilities for camera-only 3D object detection. Rethinking Temporal Fusion with a Unified Gradient Descent View, which presents a temporal fusion method for vision-based 3D semantic occupancy prediction.