A Lightweight Spatiotemporal Network for Online Eye Tracking
In our latest conference paper, an exploration of BrainChip neuromorphic techniques in event-based data processing, tailored specifically for optimizing efficiency in edge computing environments. Discover the foundational advancements driving efficiency and efficacy in event-based data processing, providing invaluable insights for researchers and practitioners in the field of edge computing.
Here’s what you’ll uncover:
– Causal Spatiotemporal Convolutional Network: A detailed examination of our thoroughly designed network, engineered to seamlessly handle event data while maximizing performance on hardware with limited resources.
– Efficiency Optimization Strategies: In-depth insights into our approach, including the deliberate simplification of architecture, strategic buffering for online inference, and achieving remarkable activation sparsity exceeding 90%.
– Affine Augmentation Strategy: An exploration of our innovative methods aimed at mitigating dataset scarcity challenges inherent in event-based systems.
– Real-world Performance Evaluation: A comprehensive analysis of our solution’s performance, demonstrated through its outstanding achievement in the AIS 2024 event-based eye tracking challenge, with a notable score of 0.9916 p10 accuracy on the Kaggle private test set.
Download the full article by filling out the form below.
