The following tutorials provide end-to-end guides on how to develop machine learning model using the MLTK:

Name Description
Keyword Spotting - On/Off Develop an ML model to detect the keywords: "on" or "off"
Keyword Spotting - Pac-Man Develop a demo to play the game Pac-Man in a web browser using the keywords: "Left", "Right", "Up", "Down", "Stop", "Go"
Keyword Spotting - Alexa Develop a demo to issue "Alexa" commands to the AVS cloud and locally play the response
Image Classification - Rock/Paper/Scissors Develop an image classification ML model to detect the hand gestures: "rock", "paper", "scissors"
Model Training in the "Cloud" Vastly improve model training times by training a model in the "cloud" using vast.ai
Logging to the Cloud Log model files and metrics to the cloud during training and evaluation using Weights & Biases
Model Optimization for MVP Hardware Accelerator Use the various MLTK tools to optimize a model to fit within an embedded device's resource constraints
Keyword Spotting with Transfer Learning Use a pre-trained model to quickly train a new model that detects the keywords: "one", "two", "three", "four"
Fingerprint Authentication Use ML to generate unique signatures from images of fingerprints to authenticate users
ONNX to TF-Lite Model Conversion Describes how to convert an ONNX formatted model file to the .tflite model format required by embedded targets
Model Debugging Describes how to debug an ML model during training
Add an Existing Script to the MLTK Describes how to convert an existing Tensorflow training script to support the MLTK training flow
Synthetic Audio Dataset Generation Describes how to generate a custom audio dataset using synthetic data. This allows for training keyword spotting ML models with custom keywords
Model Quantization Tips Provides tips on how to gain better quantization for your model
Quantized LSTM Describes how to create a quantized keyword spotting model with an LSTM layer