Created
June 13, 2020 13:38
-
-
Save antonyharfield/8398ad6fd2400e25dc8976fb2918bf1e to your computer and use it in GitHub Desktop.
YAMNet to TFLite conversion
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
import tensorflow as tf | |
from tensorflow.keras import Model, layers | |
import features as features_lib | |
import features_tflite as features_tflite_lib | |
import params | |
from yamnet import yamnet | |
def yamnet_frames_tflite_model(feature_params): | |
"""Defines the YAMNet model suitable for tflite conversion.""" | |
num_samples = int(round(params.SAMPLE_RATE * 0.975)) | |
waveform = layers.Input(batch_shape=(1, num_samples)) | |
spectrogram = features_tflite_lib.waveform_to_log_mel_spectrogram( | |
tf.squeeze(waveform, axis=0), feature_params) | |
patches = features_lib.spectrogram_to_patches(spectrogram, feature_params) | |
predictions = yamnet(patches) | |
frames_model = Model(name='yamnet_frames', | |
inputs=waveform, outputs=[predictions, spectrogram]) | |
return frames_model | |
def main(): | |
# Load the model and weights | |
model = yamnet_frames_tflite_model(params) | |
model.load_weights('yamnet.h5') | |
# Convert the model | |
converter = tf.lite.TFLiteConverter.from_keras_model(model) | |
tflite_model = converter.convert() | |
open("yamnet.tflite", "wb").write(tflite_model) | |
if __name__ == '__main__': | |
main() |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment