On the Importance of Capturing a Sufficient Diversity of Perspective for the Classification of micro-PCBs
The study compared two neural network models for classifying micro-PCBs: one using standard fully connected layers and another using Homogeneous Vector Capsules (HVCs) to model sub-component equivariance. The HVC model performed significantly better, achieving 99.06% accuracy. The research concludes that capturing a diverse set of perspectives during training is crucial for accurate classification when dealing with novel viewpoints.