Edge Impulse End-to-End Tutorials

https://docs.edgeimpulse.com/tutorials/end-to-end
Admin
Edge Impulse is a machine learning platform that enables developers to build full machine learning...
Tokens:17,643
Snippets:283
Trust Score:10
Update:2 months ago
Show doc for...
Tokens:
Raw
### View Project Summary Output

Source: https://docs.edgeimpulse.com/tutorials/end-to-end/motion-recognition-using-agent

Example output returned by the agent after querying project information and sample counts.

```bash
Project: Gesture Classifier (ID: 12345)
Training samples: 47
  idle   18
  wave   16
  punch  13
Testing samples: 0
```

--------------------------------

### Agent Build Status Output

Source: https://docs.edgeimpulse.com/tutorials/end-to-end/motion-recognition-using-agent

Example output showing the agent's progress during the deployment build process.

```text
Build started (type: arduino).
Polling...  done (34 s).
Downloaded to ./build/arduino-library.zip  (1.2 MB)
```

--------------------------------

### Inference Output Example

Source: https://docs.edgeimpulse.com/tutorials/end-to-end/sound-recognition

Example output showing DSP, classification, and anomaly detection results from the device.

```text
Starting inferencing in 2 seconds...
Recording
Recording OK
Predictions (DSP: 399 ms., Classification: 175 ms., Anomaly: 0 ms.):
    faucet: 0.03757
    noise: 0.96243
Starting inferencing in 2 seconds...
```

--------------------------------

### Monitor Data Upload Progress

Source: https://docs.edgeimpulse.com/tutorials/end-to-end/motion-recognition-using-agent

Example console output showing the agent's progress while uploading labeled training samples.

```bash
Uploading idle samples...   18/18 done
Uploading wave samples...   16/16 done
Uploading punch samples...  13/13 done
Total uploaded: 47 training samples
```

--------------------------------

### Example Inference Output

Source: https://docs.edgeimpulse.com/tutorials/end-to-end/keyword-spotting

Sample output showing classification results and timing metrics from the impulse runner.

```text
Edge Impulse impulse runner v1.9.1
[SER] Connecting to /dev/tty.usbmodem0004401658161
Predictions (DSP: 143 ms., Classification: 13 ms., Anomaly: 0 ms.):
    helloworld: 	0.98828
    noise: 	        0.0039
    unknown: 	    0.00781
```

--------------------------------

### Arduino Inference Code for Opta PLC

Source: https://docs.edgeimpulse.com/tutorials/end-to-end/temp-anomaly-detection-opta-plc

This example demonstrates how to collect sensor data, run inference using the Edge Impulse SDK, and trigger hardware actions based on anomaly detection results.

```cpp
#include <Your_Edge_Impulse_Inference_Library.h>
#include <edge-impulse-sdk/classifier/ei_run_classifier.h>

// Define the feature buffer
float features[EI_CLASSIFIER_DSP_INPUT_FRAME_SIZE];

void setup() {
  Serial.begin(115200);
  while (!Serial);

  pinMode(TEMP_SENS, INPUT);
  Serial.println("Edge Impulse Inference Example");
}

void loop() {
  // Collect data
  for (size_t i = 0; i < EI_CLASSIFIER_DSP_INPUT_FRAME_SIZE; i++) {
    features[i] = analogRead(TEMP_SENS) * (10.0 / 1024.0);
    delay(10); // Adjust delay as needed
  }

  // Prepare signal
  signal_t signal;
  numpy::signal_from_buffer(features, EI_CLASSIFIER_DSP_INPUT_FRAME_SIZE, &signal);

  // Run inference
  ei_impulse_result_t result = { 0 };
  EI_IMPULSE_ERROR res = run_classifier(&signal, &result, false);

  if (res != EI_IMPULSE_OK) {
    Serial.print("ERR: Failed to run classifier (");
    Serial.print(res);
    Serial.println(")");
    return;
  }

  // Print results
  Serial.println("Inference results:");
  for (size_t ix = 0; ix < EI_CLASSIFIER_LABEL_COUNT; ix++) {
    Serial.print(result.classification[ix].label);
    Serial.print(": ");
    Serial.println(result.classification[ix].value);
  }

  // Control heaters based on anomaly detection
  if (result.anomaly > ANOMALY_THRESHOLD) {
    // Anomaly detected, take action
    Serial.println("Anomaly detected!");
    digitalWrite(HEAT_LEFT, LOW);
    digitalWrite(HEAT_RIGHT, LOW);
  } else {
    // Normal operation
    digitalWrite(HEAT_LEFT, HIGH);
    digitalWrite(HEAT_RIGHT, HIGH);
  }

  delay(1000);
}
```

--------------------------------

### Inference Output Example

Source: https://docs.edgeimpulse.com/tutorials/end-to-end/motion-recognition

Sample output showing the results of signal processing and classification, including timing and prediction scores.

```text
Starting inferencing in 2 seconds...
Sampling... Storing in file name: /fs/device-classification261
Tensor shape: 4
Predictions (DSP: 17 ms., Classification: 1 ms., Anomaly: 0 ms.):
    idle: 0.00004
    snake: 0.00012
    updown: 0.00009
    wave: 0.99976
    anomaly score: 0.032
Finished inferencing, raw data is stored in '/fs/device-classification261'. Use AT+UPLOADFILE to send back to Edge Impulse.
```

--------------------------------

### Sample CSV Structure for HR/HRV Data

Source: https://docs.edgeimpulse.com/tutorials/end-to-end/hr-hrv-estimation

Example of the expected CSV format for physiological data, including timestamps, accelerometer axes, PPG, and heart rate labels.

```csv
Time,acc_x,acc_y,acc_z,ppg,eda,temp,hr,activity,activity_label,Subject
2018-06-29 08:44:52.000000,-27,1,58,-0.0,0.0,32.15,76.0,NO_ACTIVITY,NO_ACTIVITY,Subject_1
2018-06-29 08:44:52.031250,-27,2,59,-0.0,0.0,32.15,76.0,NO_ACTIVITY,NO_ACTIVITY,Subject_1
2018-06-29 08:44:52.062500,-27,1,59,-0.0,0.0,32.15,76.0,NO_ACTIVITY,NO_ACTIVITY,Subject_1
2018-06-29 08:44:52.093750,-27,1,58,-0.0,0.0,32.15,76.0,NO_ACTIVITY,NO_ACTIVITY,Subject_1
2018-06-29 08:44:52.125000,-27,1,59,-0.0,0.0,32.15,76.0,NO_ACTIVITY,NO_ACTIVITY,Subject_1
2018-06-29 08:44:52.156250,-27,1,59,0.0,0.832765,32.15,76.0,NO_ACTIVITY,NO_ACTIVITY,Subject_1
2018-06-29 08:44:52.187500,-27,1,59,0.01,0.832765,32.15,76.0,NO_ACTIVITY,NO_ACTIVITY,Subject_1
2018-06-29 08:44:52.218750,-27,1,58,-0.03,0.832765,32.15,76.0,NO_ACTIVITY,NO_ACTIVITY,Subject_1
2018-06-29 08:44:52.250000,-27,1,58,-0.05,0.832765,32.15,76.0,NO_ACTIVITY,NO_ACTIVITY,Subject_1
2018-06-29 08:44:52.281250,-27,1,59,0.13,0.832765,32.15,76.0,NO_ACTIVITY,NO_ACTIVITY,Subject_1
2018-06-29 08:44:52.312500,-27,1,59,0.66,0.832765,32.15,76.0,NO_ACTIVITY,NO_ACTIVITY,Subject_1
2018-06-29 08:44:52.343750,-27,1,59,1.37,0.832765,32.15,76.0,NO_ACTIVITY,NO_ACTIVITY,Subject_1
2018-06-29 08:44:52.375000,-27,1,59,2.06,0.832765,32.15,76.0,NO_ACTIVITY,NO_ACTIVITY,Subject_1
2018-06-29 08:44:52.406250,-27,1,59,2.79,1.180231,32.15,76.0,NO_ACTIVITY,NO_ACTIVITY,Subject_1
2018-06-29 08:44:52.437500,-27,1,58,3.8,1.180231,32.15,76.0,NO_ACTIVITY,NO_ACTIVITY,Subject_1
2018-06-29 08:44:52.468750,-27,1,59,5.06,1.180231,32.15,76.0,NO_ACTIVITY,NO_ACTIVITY,Subject_1
2018-06-29 08:44:52.500000,-27,1,58,6.2,1.180231,32.15,76.0,NO_ACTIVITY,NO_ACTIVITY,Subject_1
2018-06-29 08:44:52.531250,-27,1,59,7.15,1.180231,32.15,76.0,NO_ACTIVITY,NO_ACTIVITY,Subject_1
2018-06-29 08:44:52.562500,-27,1,59,8.31,1.180231,32.15,76.0,NO_ACTIVITY,NO_ACTIVITY,Subject_1
2018-06-29 08:44:52.593750,-27,1,59,10.09,1.180231,32.15,76.0,NO_ACTIVITY,NO_ACTIVITY,Subject_1
2018-06-29 08:44:52.625000,-27,1,59,11.88,1.180231,32.15,76.0,NO_ACTIVITY,NO_ACTIVITY,Subject_1
2018-06-29 08:44:52.656250,-27,2,59,13.14,1.618462,32.15,76.0,NO_ACTIVITY,NO_ACTIVITY,Subject_1
2018-06-29 08:44:52.687500,-27,1,59,14.65,1.618462,32.15,76.0,NO_ACTIVITY,NO_ACTIVITY,Subject_1
2018-06-29 08:44:52.718750,-27,1,59,17.91,1.618462,32.15,76.0,NO_ACTIVITY,NO_ACTIVITY,Subject_1
2018-06-29 08:44:52.750000,-27,1,58,22.9,1.618462,32.15,76.0,NO_ACTIVITY,NO_ACTIVITY,Subject_1
2018-06-29 08:44:52.781250,-27,1,59,27.58,1.618462,32.15,76.0,NO_ACTIVITY,NO_ACTIVITY,Subject_1
...
Download the CSV here
```

--------------------------------

### End-to-end Edge Impulse Workflow Prompt

Source: https://docs.edgeimpulse.com/tutorials/end-to-end/motion-recognition-using-agent

A comprehensive prompt designed to guide an AI agent through the Edge Impulse project lifecycle, requiring user input at each critical decision point.

```markdown
  Run an end-to-end Edge Impulse workflow for me, pausing to ask me questions at each step:

  1. Fetch my project info and show me a summary (name, ID, sample counts per label). Ask me: does this look right, or should we use a different project?

  2. Check whether I have enough training data. If not, ask me: do I want to upload files from a local folder, or generate synthetic data? Wait for my answer before proceeding.

  3. Review the training/testing split. If fewer than 20% of samples are in the test set, ask me: should we move some samples over?

  4. Start a training job. While it runs, show me the job URL so I can monitor it. When it finishes, show me the accuracy and loss. Ask me: are these results good enough, or should we adjust and retrain?

  5. Run model testing and show the confusion matrix and per-class F1 scores. Ask me: is there a class I want to improve before exporting?

  6. Ask me: what deployment format do I want? (Arduino, C++ library, TFLite, ONNX, or other) Then build and download the artifact to ./build.

  7. Summarise what was done and suggest next steps based on the results.

  Stop and wait for my input at each question. Do not skip ahead.
```

--------------------------------

### List and validate training samples

Source: https://docs.edgeimpulse.com/tutorials/end-to-end/motion-recognition-using-agent

Use this prompt to identify potential labeling inconsistencies by comparing sample IDs against filenames.

```text
List all training samples and flag any that look mislabelled in my Edge Impulse project — show me sample IDs where the label doesn't match the file name
```

--------------------------------

### Build and Run C++ SDK on Linux

Source: https://docs.edgeimpulse.com/tutorials/end-to-end/motion-recognition-using-agent

Commands to compile the Edge Impulse C++ SDK and execute inference on a sample CSV file.

```bash
cd ./build/edge-impulse-sdk
mkdir -p build && cd build
cmake .. -DCMAKE_BUILD_TYPE=Release
make -j4

# Run inference on a sample CSV:
./edge-impulse-linux-runner --run-impulse --sample ./data/idle/sample_001.csv
```

--------------------------------

### Render CardOpenStudioTutorial component

Source: https://docs.edgeimpulse.com/tutorials/end-to-end/visual-regression

Usage of the tutorial component to link to the public Edge Impulse project.

```jsx
<CardOpenStudioTutorial href="https://studio.edgeimpulse.com/public/571331/latest" />
```

--------------------------------

### Bulk-relabel training samples

Source: https://docs.edgeimpulse.com/tutorials/end-to-end/motion-recognition-using-agent

Automate the relabeling process for samples matching specific filename patterns.

```text
Relabel all training samples whose filename contains "gesture_idle" to "idle" in my Edge Impulse project
```

--------------------------------

### Execute Edge Impulse Data Forwarder

Source: https://docs.edgeimpulse.com/tutorials/end-to-end/temp-anomaly-detection-opta-plc

Command to initiate the data forwarding process from the Opta PLC to the Edge Impulse platform.

```bash
edge-impulse-data-forwarder
```

--------------------------------

### Run Impulse via CLI

Source: https://docs.edgeimpulse.com/tutorials/end-to-end/image-classification

Commands to execute the deployed model on a connected device via serial connection.

```bash
$ edge-impulse-run-impulse
```

```bash
$ edge-impulse-run-impulse --debug
```

```bash
$ edge-impulse-run-impulse --continuous
```

--------------------------------

### Run Impulse via Serial

Source: https://docs.edgeimpulse.com/tutorials/end-to-end/motion-recognition

Executes the deployed model on the connected device to perform real-time sensor sampling and classification.

```bash
$ edge-impulse-run-impulse
```

--------------------------------

### Run Edge Impulse Linux Runner

Source: https://docs.edgeimpulse.com/tutorials/end-to-end/object-detection-bounding-boxes

Command to build, download, and execute the model on a Linux-based development board.

```bash
edge-impulse-linux-runner
```

--------------------------------

### Run Continuous Inference

Source: https://docs.edgeimpulse.com/tutorials/end-to-end/motion-recognition

Executes the model in continuous sampling mode for ongoing gesture recognition.

```bash
$ edge-impulse-run-impulse --continuous
```

--------------------------------

### Configure Edge Impulse Credentials

Source: https://docs.edgeimpulse.com/tutorials/end-to-end/motion-recognition-using-agent

Set the required environment variables to authenticate the agent with your Edge Impulse project.

```md
EI_API_KEY="ei_..."
EI_PROJECT_ID="12345"
```

--------------------------------

### Arduino Data Forwarder Sketch

Source: https://docs.edgeimpulse.com/tutorials/end-to-end/temp-anomaly-detection-opta-plc

Reads temperature data from an analog pin and formats it for the Edge Impulse data forwarder.

```cpp
void setup() {
  Serial.begin(115200);
  while (!Serial);

  pinMode(TEMP_SENS, INPUT);
  Serial.println("Edge Impulse Data Forwarder Example");
}

void loop() {
  float temperature = analogRead(TEMP_SENS) * (10.0 / 1024.0); // Convert to voltage
  Serial.print("Temperature: ");
  Serial.println(temperature);

  // Send data in edge-impulse-data-forwarder format
  Serial.print("1,");
  Serial.println(temperature);

  delay(1000);
}
```

--------------------------------

### Enable HR/HRV runtime features

Source: https://docs.edgeimpulse.com/tutorials/end-to-end/hr-hrv-estimation

Use this macro to conditionally include heart rate calculation variables in your deployment.

```cpp
#if EI_DSP_ENABLE_RUNTIME_HR == 1
    ei_impulse_result_hr_t hr_calcs;
#endif
```

--------------------------------

### Automate train-test split

Source: https://docs.edgeimpulse.com/tutorials/end-to-end/motion-recognition-using-agent

Randomly move a percentage of samples from the training set to the testing set to maintain an 80/20 distribution.

```text
Move 20% of training samples for each label to the testing set in my Edge Impulse project, picking them randomly
```

--------------------------------

### Simulate Real-time Classification

Source: https://docs.edgeimpulse.com/tutorials/end-to-end/obd-automotive-data

Plays back a CSV recording to simulate real-time classification of OBD-II signals using a trained model.

```python
python3 play_csv_to_eim.py \
  --model ./obd_fault.eim \
  --csv ./airleak_MAF_NOx.csv \
  --axes "RPM [RPM],PEDAL INPUT [%],MAF [g/s],NOx [ppm]" \
  --hz 2 --window-ms 2000 --step-ms 1000 --label-col fault_label
```

--------------------------------

### Define Opta PLC Pin Constants

Source: https://docs.edgeimpulse.com/tutorials/end-to-end/temp-anomaly-detection-opta-plc

Assigns hardware pins for the user button, heaters, and temperature sensor to named constants.

```cpp
#define BTN         BTN_USER       // User button on Opta WiFi
#define HEAT_LEFT   2              // Left heater control pin
#define HEAT_RIGHT  3              // Right heater control pin
#define TEMP_SENS   A7             // Temperature sensor analog pin
```

--------------------------------

### BibTeX Citation for Automotive Diagnostics Paper

Source: https://docs.edgeimpulse.com/tutorials/end-to-end/obd-automotive-data

Use this BibTeX entry to cite the research paper on federated automotive diagnostics.

```bibtex
@inproceedings{jordan2024semantics,
  author    = {Jordan, E. and Serrano, M. and Gyrard, A. and Song, E. and Roth, T. and Wollman, D.},
  title     = {Semantics for Enhancing Communications- and Edge-Intelligence-enabled Smart Sensors: A Practical Use Case in Federated Automotive Diagnostics},
  booktitle = {IECON 2024 - 50th Annual Conference of the IEEE Industrial Electronics Society},
  address   = {Chicago, IL, USA},
  year      = {2024},
  pages     = {1--6},
  doi       = {10.1109/IECON55916.2024.10905565}
}
```

--------------------------------

### Test Heater Control

Source: https://docs.edgeimpulse.com/tutorials/end-to-end/temp-anomaly-detection-opta-plc

Toggles the heater pins high and low to verify physical activation of the connected hardware.

```cpp
void setup() {
  pinMode(HEAT_LEFT, OUTPUT);
  pinMode(HEAT_RIGHT, OUTPUT);
}

void loop() {
  digitalWrite(HEAT_LEFT, HIGH);
  digitalWrite(HEAT_RIGHT, HIGH);
  delay(1000);
  digitalWrite(HEAT_LEFT, LOW);
  digitalWrite(HEAT_RIGHT, LOW);
  delay(1000);
}
```

--------------------------------

### Access Live Classification Feed

Source: https://docs.edgeimpulse.com/tutorials/end-to-end/object-detection-bounding-boxes

URL output provided by the runner to view camera feed and classification results in a browser.

```text
Want to see a feed of the camera and live classification in your browser? Go to http://192.168.1.19:4912
```

--------------------------------

### Test Temperature Sensor Reading

Source: https://docs.edgeimpulse.com/tutorials/end-to-end/temp-anomaly-detection-opta-plc

Reads analog values from the temperature sensor and outputs them to the Serial Monitor for verification.

```cpp
void setup() {
  Serial.begin(9600);
  pinMode(TEMP_SENS, INPUT);
}

void loop() {
  int sensorValue = analogRead(TEMP_SENS);
  Serial.println(sensorValue);
  delay(250);
}
```

--------------------------------

### Update Inference Sketch with Shared Variable

Source: https://docs.edgeimpulse.com/tutorials/end-to-end/temp-anomaly-detection-opta-plc

Define a global variable to store inference results, allowing the PLC ladder logic to access the data produced by the sketch.

```cpp
float anomaly_score;  // Shared variable to store the inference result

void setup() {
  Serial.begin(115200);
  pinMode(A7, INPUT);  // Sensor input
}

void loop() {
  float sensorValue = analogRead(A7) * (10.0 / 1024.0);  // Read sensor value
  anomaly_score = run_inference(sensorValue);  // Store inference result
  delay(1000);  // Adjust as needed
}
```

--------------------------------

### CSV Header Format

Source: https://docs.edgeimpulse.com/tutorials/end-to-end/obd-automotive-data

Required header structure for OBD-II data files, including time, label, and signal columns.

```text
time (ms.),fault_label,RPM [RPM],PEDAL INPUT [%],MAF [g/s],NOx [ppm]
```
Tokens:
Raw