search

Ant-Brain / EfficientWord-Net

OneShot Learning-based hotword detection.
https://ant-brain.github.io/EfficientWord-Net/
Apache License 2.0
215 stars 34 forks source link

baseNetwork and Siamese modelCheckpoints used in train script #44

Closed artem-tok closed 8 months ago

artem-tok commented 8 months ago

Hello! In script https://github.com/Ant-Brain/EfficientWord-Net/blob/main/training.ipynb you are referring to baseNetwork = tf.keras.models.load_model("/content/drive/MyDrive/Siamese/modelCheckpoints_old/model-8-01-0.96.h5").get_layer("basemodel") and 

model = tf.keras.models.load_model(
    "/content/drive/MyDrive/Siamese/modelCheckpoints/model-4-15-loss_0.072_val_acc0.947.h5",
    custom_objects={"triplet_loss":triplet_loss})

on your drive. Could you please share this models?

TheSeriousProgrammer commented 8 months ago

That model is old and very poor interms of accuracy, consider using the latest resnet-arcloss based ones its built with pytorch

TheSeriousProgrammer commented 8 months ago

Pytorch Training Repo https://github.com/Ant-Brain/EfficientWord-Net-Trainer, simple_trainer.py is the entry point

artem-tok commented 8 months ago

@TheSeriousProgrammer many thanks!)