7. Examples & Applications
tinyml-modelzoo ships ready-to-run examples under examples/. Each example folder has one or more
config YAML files (one per target device or variant), and most have a readme.md. This section has one
page per example, grouped by task.
Note
Many example folders have no plain config.yaml. Use the config file names listed on each page,
for example examples/dc_arc_fault/config_dsk.yaml.
7.1. Running an Example
From the tinyml-modelzoo folder:
./run_tinyml_modelzoo.sh examples/generic_timeseries_classification/config.yaml
run_tinyml_modelzoo.bat examples\generic_timeseries_classification\config.yaml
./run_tinyml_modelzoo.ps1 examples/generic_timeseries_classification/config.yaml
Results are written to ./data/projects/<project_name>/run/<date-time>/<model_name>/; see
Understanding the Run Output. If you installed the wheel only, see
Wheel Installation for how to get the example configs. To adapt an example, see
How to Create a ModelZoo Config YAML. To run a trained example on a board, see Deploying a ModelZoo Run to an SDK Example (Golden-Vector Check).
Important
Result numbers on the example pages come from the example readme.md files (written for earlier
releases) or from short automated test runs of 1.5.0, as labelled on each page. They show what to
expect, not a benchmark.
7.2. Generic Examples (Start Here)
One per timeseries task. Use these to learn the flow, then point the config at your own data.
Example |
Description |
|---|---|
Hello World: classify sine, square and sawtooth waveforms. |
|
Hello World regression on synthetic data. |
|
Hello World forecasting on a simulated thermostat signal. |
|
Autoencoder anomaly detection on synthetic data. |
7.3. Timeseries Classification
Example |
Description |
|---|---|
DC arc fault detection from current waveforms (classification, anomaly detection, on-device learning). |
|
AC arc fault detection on MSPM0G5187. |
|
Motor bearing fault from 3-axis vibration (classification and anomaly detection). |
|
Blower imbalance from a 3-variable signal. |
|
Fan blade faults from an ADXL355 accelerometer, including wireless MCUs and on-device training. |
|
Transmission line fault classification (2-class and 6-class). |
|
Smart grid stable vs unstable (UCI 4-node star grid). |
|
Single-phase on-board-charger grid faults on F29H85. |
|
Gas identification from 10 MOS sensors. |
|
Human activity recognition (WISDM) with a residual model. |
|
ECG classification (classification and anomaly detection). |
|
Appliance usage classification (ESDA NILM and PLAID). |
|
Motion detection from a PIR sensor on MSPM0 and wireless MCUs. |
|
Gearbox fault from 4-channel vibration. |
|
Fall detection from an accelerometer (SisFall). |
|
Hand gestures from a 3-axis accelerometer. |
|
Presence detection from Wi-Fi CSI on CC35X1 (new in 1.5.0). |
7.4. Timeseries Regression
Example |
Description |
|---|---|
PMSM torque estimation. |
|
Induction motor speed (RPM) prediction. |
|
Laundry load weighing from motor signals. |
|
MOSFET case-temperature prediction (linear model plus AI residual). |
|
Battery state-of-charge estimation (new in 1.5.0). |
7.5. Timeseries Forecasting
Forecasting Examples Guide lists all forecasting configs, the windowing rules and the FCST_* models.
Example |
Description |
|---|---|
PMSM rotor magnet temperature, one step ahead. |
|
Indoor temperature forecasting for HVAC. |
7.6. Anomaly Detection
Anomaly detection configs live inside several example folders. Anomaly Detection Examples Overview lists all of them with thresholds and troubleshooting.
7.7. Audio Classification
Example |
Description |
|---|---|
12-class keyword spotting with MFCC features. |
|
Wake word detection with a learned filterbank (new in 1.5.0). |
|
Glass break detection (new in 1.5.0). |
|
Cough detection (new in 1.5.0). |
7.8. Image Classification
Image Classification Examples Guide is an overview of the image examples.
Example |
Description |
|---|---|
Digit recognition with LeNet-5. |
|
QR code vs barcode vs other. |
|
Coffee roast level classification. |
7.9. Radar
Example |
Description |
|---|---|
Pose and fall classification from radar point clouds (new in 1.5.0). |