edgemodelkit 2.0.0

edgemodelkit: A Python library for seamless sensor data acquisition and logging.

EdgeModelKit: Sensor Data Acquisition and Logging Library

EdgeModelKit is a Python library developed by EdgeNeuron, designed to simplify sensor data acquisition, logging, and real-time processing for IoT devices. It works seamlessly with the DataLogger script from the EdgeNeuron Arduino library, and now supports HTTP-based acquisition for devices that expose REST APIs.

---

Features

• Serial Communication: Acquire data from devices connected via serial ports.
• HTTP Communication: Fetch data directly from devices exposing REST APIs.
• Flexible Data Fetching: Retrieve sensor data as Python lists or NumPy arrays.
• Customizable Logging: Log sensor data into CSV files with optional timestamps and counters.
• Class-Based Organization: Log data with class labels to prepare datasets for machine learning tasks.
• Custom Preprocessing: Apply custom preprocessing functions to sensor data before logging or inference.
• Error Handling: Gracefully handles data decoding errors, missing keys, or network errors.

---

Usage Prerequisites

This library is designed to work with devices that provide structured JSON sensor data either:

1. Over Serial, using the DataLogger script in the EdgeSense Arduino library.
2. Over HTTP, where the device exposes a REST API returning JSON sensor data.

---

Installation

pip install edgemodelkit

---

Quick Start

1. Initialize the DataFetcher

Serial Mode:

from edgemodelkit import DataFetcher

# Initialize for Serial communication
fetcher = DataFetcher(source="serial", serial_port="COM3", baud_rate=9600)

HTTP Mode:

from edgemodelkit import DataFetcher

# Initialize for HTTP communication
fetcher = DataFetcher(source="http", api_url="http://192.168.26.123")

---

2. Fetch Sensor Data

# Fetch data as a Python list
sensor_data = fetcher.fetch_data(return_as_numpy=False)
print("Sensor Data:", sensor_data)

# Fetch data as a NumPy array
sensor_data_numpy = fetcher.fetch_data(return_as_numpy=True)
print("Sensor Data (NumPy):", sensor_data_numpy)

---

3. Log Sensor Data

# Log 10 samples with timestamp and count columns
fetcher.log_sensor_data(class_label="ClassA", num_samples=10, add_timestamp=True, add_count=True)

CSV files are saved under a folder named Dataset, with subfolders organized by class_label.

---

CSV Logging Details

The generated CSV file is named after the sensor (e.g., TemperatureSensor_data_log.csv) and includes:

• Timestamp (optional)
• Sample Count (optional)
• Data Columns (data_value_1, data_value_2, …)

---

Real-Time Data Processing Example

from edgemodelkit import DataFetcher

# Works with both Serial and HTTP
fetcher = DataFetcher(source="http", api_url="http://192.168.26.123")

def custom_preprocess(data):
    # Example: Normalize the data
    return (data - min(data)) / (max(data) - min(data))

try:
    while True:
        sensor_data = fetcher.fetch_data(return_as_numpy=True)
        print("Received Data (Raw):", sensor_data)

        processed_data = custom_preprocess(sensor_data)
        print("Preprocessed Data:", processed_data)

        # prediction = model.predict(processed_data)
        # print("Prediction:", prediction)
finally:
    fetcher.close_connection()

---

Using ModelPlayGround

1. Initialize and Load Model

from edgemodelkit import ModelPlayGround

playground = ModelPlayGround()
playground.load_model(model_path="path_to_your_model.keras")

2. Model Summary and Stats

playground.model_summary()
playground.model_stats()

3. Convert Model to TFLite

playground.model_converter(quantization_type="default")
playground.model_converter(quantization_type="float16")
playground.model_converter(quantization_type="int8")

4. Test TFLite Model on Live Data

from edgemodelkit import DataFetcher

fetcher = DataFetcher(source="serial", serial_port="COM3", baud_rate=9600)

def custom_preprocess(data):
    return (data - min(data)) / (max(data) - min(data))

playground_output = playground.edge_testing(
    data_fetcher=fetcher,
    preprocess_func=custom_preprocess
)
print("Model Prediction:", playground_output['ModelOutput'])
print("Sensor data: ", playground_output['SensorData'])

5. Test with an Existing TFLite Model

playground_output = playground.edge_testing(
    tflite_model_path="path_to_tflite_model.tflite",
    data_fetcher=fetcher,
    preprocess_func=custom_preprocess
)
print("Model Prediction:", playground_output['ModelOutput'])
print("Sensor data: ", playground_output['SensorData'])

---

Disclaimer

Currently, the ModelPlayGround class supports .keras models for conversion and testing. Support for additional formats may be added in future updates.

---

Contributing

We welcome contributions! Submit bug reports, feature requests, or pull requests at GitHub.

---

License

MIT License. See the LICENSE file for details.

---

Support

📧 [e-mail] 🌐 GitHub Repository

---

About EdgeNeuron

EdgeNeuron is a pioneer in edge computing solutions, enabling developers to build intelligent IoT applications with state-of-the-art tools and libraries. Learn more at edgeneuronai.com.

---

Do you want me to also include a **feature comparison table (Serial vs HTTP)** in this README to make the new dual-source support stand out?