Extended Data Fig. 2: Influence of the noise-level estimator on the activation of hidden layers. | Nature Machine Intelligence

Extended Data Fig. 2: Influence of the noise-level estimator on the activation of hidden layers.

From: Self-supervised deep learning for tracking degradation of perovskite light-emitting diodes with multispectral imaging

Extended Data Fig. 2: Influence of the noise-level estimator on the activation of hidden layers.The alternative text for this image may have been generated using AI.

Activation of the first 32 channels from the model output of the first BN layer using noisy image inputs of various noise levels. a, PA-CNN. b, PA-Net. The activation in b varies across different noise levels, which is desired by our proposed model. The activation features were extracted using methods described in Supplementary Note 4.

Source data

Back to article page