A rough time comparison needs to be performed to XRFI as it exists in RTP. XRFI takes ~20 minutes on a standard observation. The majority of time for the FCN to predict appears to be in loading the model into memory and initializing the weights?
A quick prediction test on 1450 waterfall visibilities (Not checked for accuracy of predictions!) including the loading of the model, folding, and unfolding of the waterfalls gives a total time of ~72s.
A rough time comparison needs to be performed to XRFI as it exists in RTP. XRFI takes ~20 minutes on a standard observation. The majority of time for the FCN to predict appears to be in loading the model into memory and initializing the weights?