Architecture of the SUNet neural network (IMAGE)

Nuclear Science and Techniques

Architecture of the SUNet neural network

Caption

SUNet comprises three core modules: shallow feature extraction (via 3×3 convolutional layer), Unet feature extraction (integrating Swin Transformer blocks and patch merging layers), and reconstruction (with Bilinear + PixelShuffle dual upsampling). Skip connections fuse multi-scale features to minimize information loss, enabling efficient denoising of noisy EPID TD images through a streamlined workflow.

Credit

Tao Qiu

Usage Restrictions

Credit must be given to the creator. Only noncommercial uses of the work are permitted.

License

CC BY-NC

Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.