Optimizing rPPG Networks: Applying Small Training Datasets to obtain Compact Deep Neural Networks

This study presents a method to enhance remote Photoplethysmography (rPPG) networks by pruning, leading to small yet dense models that perform effectively even with limited training samples. The approach focuses on reducing the network's complexity while maintaining accuracy, making it suitable for real-time applications. Experimental results demonstrate that the pruned networks achieve competitive performance compared to larger counterparts, highlighting their potential for efficient deployment in resource-constrained environments.