Fig. 2: Architecture of CenterNet detector in EPicker. | Nature Communications

Fig. 2: Architecture of CenterNet detector in EPicker.

From: EPicker is an exemplar-based continual learning approach for knowledge accumulation in cryoEM particle picking

Fig. 2

The feature extraction sub-network is a cascade of a convolutional network (green) and a deconvolutional network (red) to extract features. Both the convolutional and the deconvolutional networks are a combination of Convolution-Batch Normalization46-ReLU(Rectified Linear Unit)47 blocks. The object location sub-network (blue) of the detector generates the heatmaps of particle center and size. The heatmap regression network is a combination of Convolution-ReLU-Convolution blocks.

Back to article page