Fig. 2From: Sfcnn: a novel scoring function based on 3D convolutional neural network for accurate and stable protein–ligand affinity predictionArchitectures specifics. a Details of the CNN1 architecture. b Details of the CNN2 architecture. c Details of the Res3 architecture based on ResNet. d Details of the Dense4 architecture based on DenseNet. Abbreviations are defined as: Conv3D (number of channels, kernel size, stride): 3D convolutional neural network layer, BN: Batch Normalization layer, ReLU: Rectified Linear Unit activation layerBack to article page