🧠 soundKIT Keyword Spotting (KWS) Tutorial¶

This tutorial walks you through using the soundKIT CLI for KWS tasks including data preparation, training, evaluation, TFLite export, and embedded deployment.

⚙️ Setup¶

In [ ]:

# Set working directory to project root (one level up)
import os
os.chdir("..")
print("✅ Working directory set to project root:", os.getcwd())

# Set working directory to project root (one level up) import os os.chdir("..") print("✅ Working directory set to project root:", os.getcwd())

In [ ]:

# Clone the soundKIT repo
!git clone https://github.com/AmbiqAI/soundkit.git
%cd soundkit

# Clone the soundKIT repo !git clone https://github.com/AmbiqAI/soundkit.git %cd soundkit

In [ ]:

# Install required dependencies
!pip install -e .

# Install required dependencies !pip install -e .

🔧 Configuration¶

In [ ]:

# Show editable KWS training config
config_path = "configs/kws.yaml"
with open(config_path, "r") as f:
    print(f.read())

# Show editable KWS training config config_path = "configs/kws.yaml" with open(config_path, "r") as f: print(f.read())

📁 Step 1: Prepare Data¶

In [ ]:

!soundkit -t kws -m data -c configs/kws.yaml

!soundkit -t kws -m data -c configs/kws.yaml

🎯 Step 2: Train Model¶

In [ ]:

!soundkit -t kws -m train -c configs/kws.yaml

!soundkit -t kws -m train -c configs/kws.yaml

To monitor training with TensorBoard:

In [ ]:

!soundkit -t kws -m train --tensorboard -c configs/kws.yaml

!soundkit -t kws -m train --tensorboard -c configs/kws.yaml

In [ ]:

# Run this in another terminal:
# tensorboard --logdir=soundkit/tasks/kws/tensorboard/your_run

# Run this in another terminal: # tensorboard --logdir=soundkit/tasks/kws/tensorboard/your_run

📊 Step 3: Evaluate Model¶

In [ ]:

!soundkit -t kws -m evaluate -c configs/kws.yaml

!soundkit -t kws -m evaluate -c configs/kws.yaml

📤 Step 4: Export Model to TFLite¶

In [ ]:

!soundkit -t kws -m export -c configs/kws.yaml

!soundkit -t kws -m export -c configs/kws.yaml

🚀 Step 5: Deploy to Embedded EVB (e.g., Apollo)¶

In [ ]:

!soundkit -t kws -m demo -c configs/kws.yaml demo.platform=evb # For EVB demo, ensure you have a microphone connected
!soundkit -t kws -m demo -c configs/kws.yaml demo.platform=pc # For PC demo, ensure you have a microphone connected

!soundkit -t kws -m demo -c configs/kws.yaml demo.platform=evb # For EVB demo, ensure you have a microphone connected !soundkit -t kws -m demo -c configs/kws.yaml demo.platform=pc # For PC demo, ensure you have a microphone connected

✅ Notes¶

• Edit kws.yaml to adjust model hyperparameters, features, and output paths
• Log files and detection visualizations are saved under the configured test and training folders