Test code for running inferences using a single model.
Code for running a consensus inference. Not all parameters have been exposed, but this script produces a raw output (pre-aggregatation for consensus) and a mean consensus result. Each frame is flipped and run across multiple models for consensus modalities.
Example export code for editing and merging multiple graphs for a more efficient consensus execution.
Code for running a consensus inference using an exported graph. This code is highly specific to the exported graph and should be used as a model for how to run inference on an exported graph.
Plots the activations of a network in addition to the final consensus onto a new video via a colormap for certainty. This includes the final selected point on the ROC curve (of 0.4811055 threshold with a 46 frame rolling average). This file plots multiple consensus approaches without a rolling average.
Plots the consensus modalities onto a new video via a probability bar on the side (color = binary above/below threshold, size = relative probability).
Generalized function for reading appended NPY output files.
Example code for running an inference on the training/validation dataset.