Fig. 1: The gradient-weighted class activation mapping (Grad-CAM) and guided Grad-CAM interpretation of CNN classifier trained by the raw data. | npj Clean Water

Fig. 1: The gradient-weighted class activation mapping (Grad-CAM) and guided Grad-CAM interpretation of CNN classifier trained by the raw data.

From: Attention improvement for data-driven analyzing fluorescence excitation-emission matrix spectra via interpretable attention mechanism

Fig. 1

a Class 0: a 3D-EEM spectrum contains three fluorescent components; b Class 1: a 3D-EEM spectrum contains four fluorescent components; c Class 2: a 3D-EEM spectrum contains five fluorescent components. Three 3D-EEM spectra containing 3–5 components were randomly selected from the training dataset.

Back to article page