Figure 4
From: Deep learning data augmentation for Raman spectroscopy cancer tissue classification
1-D convolution process. The convolution kernel (filter) performs a convolution operation on each location of the input signal \({\mathbf {x}}\), where a window with size k (the dashed rectangular box) is used to extract local signal for analysis. The result of each convolution calculation outputs a point as shown in the feature map (the lower panel). The convolution kernel will slide, from left to right, through the signal \({\mathbf {x}}\) to generate feature map.
