Leo

By TheRealKANi Inspired by: Leo

Trained on Google Colab

hey_leo.tflite

This model reacts to Hey Leo.

Training

number_of_examples = 40000
number_of_training_steps = 20000
false_activation_penalty = 1500

Results

Final Model Accuracy: 0.6293749809265137
Final Model Recall: 0.25975000858306885
Final Model False Positives per Hour: 0.0