Static buffer inference on Android

This tutorial shows how to run an Edge Impulse model on Android using hardcoded test data with no sensors required. It’s a useful starting point before moving on to camera, audio, or motion input.

What you’ll build

A basic Android app that:

Loads a C++ model via Android NDK
Runs inference on static test features
Displays classification results

Prerequisites

Trained Edge Impulse model
Android Studio with NDK and CMake installed
Basic familiarity with Android development

1. Clone the repository

git clone https://github.com/edgeimpulse/example-android-inferencing.git
cd example-android-inferencing/example_static_buffer

2. Download TensorFlow Lite libraries

cd app/src/main/cpp/tflite

# Windows
download_tflite_libs.bat

# macOS/Linux
sh download_tflite_libs.sh

3. Export your model

In Edge Impulse Studio, go to Deployment
Select Android (C++ library)
Click Build and download the .zip

4. Integrate the model

Extract the downloaded .zip file

Copy all files to:

example_static_buffer/app/src/main/cpp/

Your structure should look like:

app/src/main/cpp/
├── edge-impulse-sdk/
├── model-parameters/
├── tflite-model/
├── native-lib.cpp
└── CMakeLists.txt 

5. Add test features

In Studio, go to Model testing
Click on a test sample
Copy the raw features
Paste into native-lib.cpp:

// app/src/main/cpp/native-lib.cpp
std::vector<float> raw_features = {
    // Paste your raw features here
    -1.2, 0.5, 2.1, ...
};

6. Build and run

Open the project in Android Studio
Build → Make Project
Run on a device or emulator

You should see classification results on screen.

How it works

Native inference

// native-lib.cpp
extern "C" JNIEXPORT jstring JNICALL
Java_com_example_test_1cpp_MainActivity_runInference(
    JNIEnv* env, jobject /* this */) {
    
    // Create signal from raw features
    signal_t signal;
    signal.total_length = raw_features.size();
    signal.get_data = &get_feature_data;
    
    // Run classifier
    ei_impulse_result_t result = {0};
    EI_IMPULSE_ERROR res = run_classifier(&signal, &result, false);
    
    // Format results
    return env->NewStringUTF(format_results(result).c_str());
}

Java/Kotlin bridge

// MainActivity.kt
class MainActivity : AppCompatActivity() {
    private external fun runInference(): String
    
    override fun onCreate(savedInstanceState: Bundle?) {
        super.onCreate(savedInstanceState)
        
        // Load native library
        System.loadLibrary("test_cpp")
        
        // Run inference and display
        val result = runInference()
        resultTextView.text = result
    }
}

Troubleshooting

Build fails with 'undefined reference to run_classifier'

Cause: Model files not copied correctlySolution:

Ensure all folders (edge-impulse-sdk, model-parameters, tflite-model) are in app/src/main/cpp/
Don’t replace the existing CMakeLists.txt

App crashes on launch

Cause: Native library not loadedSolution:

Verify System.loadLibrary("test_cpp") matches your library name
Check Build output for compilation errors

Wrong classification results

Cause: Test features don’t match model inputSolution:

Copy features from Studio’s Model Testing page
Ensure feature count matches model input size
Check feature order (x, y, z for accelerometer, etc.)

OVERVIEW

END-TO-END

TOPICS

TOOLS

INTEGRATIONS

What you’ll build

Prerequisites

1. Clone the repository

2. Download TensorFlow Lite libraries

3. Export your model

4. Integrate the model

5. Add test features

6. Build and run

How it works

Native inference

Java/Kotlin bridge

Troubleshooting

Next steps

​What you’ll build

​Prerequisites

​1. Clone the repository

​2. Download TensorFlow Lite libraries

​3. Export your model

​4. Integrate the model

​5. Add test features

​6. Build and run

​How it works

​Native inference

​Java/Kotlin bridge

​Troubleshooting

​Next steps

What you’ll build

Prerequisites

1. Clone the repository

2. Download TensorFlow Lite libraries

3. Export your model

4. Integrate the model

5. Add test features

6. Build and run

How it works

Native inference

Java/Kotlin bridge

Troubleshooting

Next steps