Skip to content

4.1.5 Real-time Result Push

1. Function Overview

Real-time Result Push is another Model Deploy method. Unlike exporting the model to run on hardware, real-time result push performs inference on the platform and instantly transmits the prediction results to the SIoT platform or external devices. In this way, users can experience the model’s performance in real-world applications without loading the full model locally or on hardware.

img

Features

  • Lightweight: The hardware does not need to perform inference computations and only receives results pushed from the platform, reducing performance requirements on the device.
  • Real-time: Model outputs are pushed instantly, facilitating interaction and rapid verification.
  • Flexible: Can interact with different terminals, such as visualization panels, smart cars, or IoT devices.
  • Easy to Use: No complex deployment operations are required, making it suitable for teaching experiments and rapid prototyping.

2. Application Scenarios

  • Teaching Demonstration: Students can visually observe the model’s recognition results and interact with the hardware.
  • Prototype Verification: Quickly verify model feasibility in the early stages of a project without waiting for full deployment.
  • IoT Integration: Combine with smart hardware, e.g., trigger car movement or lighting control based on recognition results.
  • Data Visualization: Display inference results in real time on the platform or terminal for analysis and demonstration.

3. Operation Guide

This guide focuses on demonstrating the workflow and specific steps for real-time result push of Image Classification on the UNIHIKER K10. The following operation steps are mainly for real-time result push after the model validation meets the expected results. For detailed steps on model training, please refer to the corresponding model Operation Guide.

3.1 Hardware List

  • Hardware Preparation
img img
UNIHIKER K10 USB数据线
  • Hardware Connection

img

3.2 Environment Preparation

Real-time result push relies on MQTT for wireless network transmission, so the actual IP address of the computer needs to be entered. The operation steps are as follows: first, start the data transmission tool (run the .bat batch file), then obtain the computer’s IP address.

Do not close the transmission tool while it is running.

  • SIoT Tool Download Links：

Under development, stay tuned.

img

img

3.3 Real-Time Result Push Application

  • Real-Time Result Push

    • After the model validation is completed and meets the requirements, click “Real-Time Result Push.” In the “Real-Time Push Server Settings,” change the MQTT server address to the local IP address (can be viewed in the Data Transmission Tool).

img

  • Once the server connection is successful, the Real-Time Result Push button turns green.

img

  • Programming

    • Add Main Controller: Click “Extensions,” then in “Board,” download and add “UNIHIKER K10.”

img

  • Add Extension Libraries: Click “Extensions,” then in the search box of “Model,” enter “Wi-Fi” and “MQTT” separately, download and add them.

img

  • Program Example

Receive real-time result messages pushed to MQTT and analyze the MQTT messages. If the MQTT message is "happy," the RGB light on the UNIHIKER K10 will display yellow; if the MQTT message is "angry," the RGB light on the UNIHIKER K10 will display red.

img

  • Test & Run
Real-time Result Push UNIHIKER K10 Execution Result
img img
img img