My book
A demo of android key word spoting based on tensorflow tutial example.
This is a very good example of using tensorflow in android APP. This example contains only keyword spoting.
The directory speech_commands is get
To train the model: python tensorflow/examples/speech_commands/train.py --wanted_words=house
To freeze pb: python tensorflow/examples/speech_commands/freeze.py --start_checkpoint=/tmp/speech_commands_train/conv.ckpt-18000 --out_file=/tmp/my_frozen_graph.pb ----wanted_words=house
After pb file generated, move to assets directory of android, android directory is an android studio project. The file used by my project is conv_actions_house_labels.txt and my_house_frozen_graph.pb.
If you want to downlad image pb file, you may change download-models.gradle file.
I think this is a good reference for anyone who want to know how tensorflow can be used by speech and android.
For more information, see Readme of each directory.