Self-attention based neural network for full-waveform echo decomposition
Due to the constraints imposed by confidentiality regulations, we are sorry that it is not feasible to fully disclose the coding, pretrained weights, and the graphical user interface (GUI) form of FW-GenTools in this Repository.
But we present the PyTorch implementations of both the classification and decomposition modules within the AFD-Net architecture. Additionally, we provide the core code from FW-GenTools that is employed in the generation of individual simulated echo signals.
Based on this code, readers are encouraged to attempt the construction of full-waveform datasets, employing the AFD-Net to first ascertain the number of signals, and subsequently decompose the waveform signals.