Java Deep Learning Projects

Restoring the trained model and inferencing

Now that we have seen how our model performed, it would be worth exploring the feasibility of restoring the already trained model. In other words, we will restore the trained model and evaluate the network performance on both validation and test sets:

private static final String TRAINED_PATH_MODEL = "bin/CatvsDog_VG16_TrainedModel_Epoch100_v1.zip";ComputationGraph computationGraph = ModelSerializer.restoreComputationGraph(new File(TRAINED_PATH_MODEL));VG16CatvDogEvaluator().runOnTestSet(computationGraph);VG16CatvDogEvaluator().runOnValidationSet(computationGraph);

In the previous line, of code, first, we restored the trained model from the disk; then we performed the evaluation on both the test ...