Fig. 4From: Three-dimensional branch segmentation and phenotype extraction of maize tassel based on deep learningOverview of the network model of maize tassel segmentation. The encoder part consists of three SA layers. Each SA layer sets a different number of sampling points, sampling radius, and Multi-Layer Perceptron (MLP) layer size. The decoder part consists of three IP layers. Each IP layer connects the features extracted by the corresponding SA layer and sets a different MLP layer sizeBack to article page