To get the prediction / output of my pre-trained model; the model predicts a symbol for each frame (column) of the convoluted image and it is necessary to conduct post-proce