Fig. 6 | Scientific Reports

Fig. 6

From: Bilateral enhancement network with signal-to-noise ratio fusion for lightweight generalizable low-light image enhancement

Fig. 6

Comparison of activation feature heatmaps for underexposed and overexposed images using ReLU and NegReLU activation functions. When the network activates features of underexposed images, ReLU tends to activate the underexposed parts, whereas NegReLU, in contrast to ReLU, tends to activate the overexposed features. Furthermore, ReLU and NegReLU exhibit similar tendencies in activating features of both underexposed and overexposed images.

Back to article page