Object Detection with Edge Impulse Using Mobile Phone

Introduction

In this project, the Object Detection project is constructed utilising the Edge Impulse platform, wherein the primary device is a mobile phone. Using the camera functionality of the mobile phone device, the project focuses on gathering data by capturing objects through the phone's camera.

Hardware Requirement

• Mobile phone

Project Development

To create a project using Edge Impulse, seven stages need to be considered:

1. Device

• Scan QR code to connect your mobile phone device to Edge Impulse

• After scanning, the website will take you to the specified page. Make sure you see the message "Connected as..." to confirm that your mobile device is successfully linked to Edge Impulse.

2. Data Acquisition

• Select the camera sensor type from the list and adjust the sample length on the same page.

• Clicking "Start sampling" records data and then press "Capture" on your mobile phone to capture the object. In this case, objects like "LED and ultrasonic sensor" are considered.

• Change the "name" to modify the data type, and then the data collected is uploaded and saved to the same data acquisition page.

• After gathering the data, select the "Labelling queue" option to frame and label the captured object.

3. Impulse Design

• You can display data graphically in charts or tables. Start by selecting and setting up impulses or features from the recommended list. Create two block models: the processing block and the learning block. (The quantity of each block model depends on what you need)

Note:  Remember to save the created impulse

• Each created impulse must be trained independently.

4. Retrain Model

• The model can be retrained using the retrain model feature with specified parameters.

5. Live Classification

• The live classification category enables users to use their phones to capture objects for the purpose of gathering and classifying test data by clicking "Start sampling".

• After capturing the test data, go back to the "Labelling Queue" page to frame and label the captured object.

6. Model Testing

• Model testing visually represents test data with the accuracy of the building model. After clicking "Classify all," you can view the output on the provided result.

7. Deployment

• Launch your building model on the mobile phone device by scanning the QR code.

• Capture various objects actively during sampling, and the results will appear on the same page.