Study on Optimal Generative Network for Synthesizing Brain Tumor-Segmented MR Images

Figure 10

The architecture of our generative model using U-Net. We use concat in the decoding process to perform a skip connection that passes important information directly from the encoder to the decoder.