Fig. 2 | Light: Science & Applications

Fig. 2

From: Deep learning enhanced light sheet fluorescence microscopy for in vivo 4D imaging of zebrafish heart beating

Fig. 2

The architecture of UI-Trans network. Each transformer block consists of two parallel Multiheaded Self-Attention (MSA) modules. Each MSA module is constructed with an embedding layer, a layer norm layer, a multi-head attention layer, another layer norm layer and a multilayer perceptron layer. In the architecture of the transformer block, the channel dimension undergoes a doubling, whereas the dimension is subjected to a reduction by a factor of two through the process of down-sampling. Cat – concatenation, Conv – convolutional, MLP – multilayer perceptron

Back to article page