Project Problem Statement

        Machine learning offers a versatile toolkit for a wide range of applications. In the context of our project, our client wanted to develop a system capable of tracking the movement of an individual's pupils. The primary objective of this endeavor is to facilitate the assessment of a user's emotional state by analyzing the movement of their pupils.
        To achieve this, three algorithms has to be ran: image preprocessing, blink detection, and eye tracking. This project is a continuation of many teams, previous teams successfully executed these algorithms independently on the Kria board. Our challenge, however, is to design a comprehensive system capable of concurrently operating all three algorithms on the Kria board, along with any other supporting algorithms required.
        This task is complicated by the hardware constraints of the Kria board, which, at the moment, is equipped with a single Deep Learning Processing Unit (DPU). Both the blink detection and eye tracking algorithms rely on the DPU. On top of that, by running multiple resource heavy algorithms at the same time, other resources, such as memory and processor cores, has to be delicately managed.
        Our objective is to design a system that optimizes the distribution of these resources to achieve a target throughput of 200 frames per second (FPS). This rate is deemed the minimal requisite for the system to effectively fulfill its intended function.