SOPHON ACADEMY / Task
Introduction

Introduction

  • Objective

This task is designed to challenge developers to utilize the processing capabilities of the Milk-V Duo 256MB version to develop a real-time video stream human figure detection system. This system will be capable of identifying and tracking human figures in a video, suitable for scenarios such as security monitoring, crowd statistics, and interactive art installations.

 

  • Detailed Task Description

1. Human Figure Detection Algorithm Development
a. Developers need to research and implement a human figure detection algorithm suitable for the hardware characteristics of the Milk-V Duo.
b. The algorithm should accurately identify and track human figures in a real-time video stream.
c. Support human figure detection in at least two different scenarios (indoor and outdoor).

 

2. Video Stream Processing
a. Implement real-time capture, processing, and display functions for video streams.
b. Ensure minimal latency during video stream processing to maintain a smooth user experience.

 

  • Performance Requirements
  1. The human figure detection algorithm should achieve real-time processing standards on the Milk-V Duo 256MB version, i.e., processing time per frame should not exceed 30 milliseconds.
  2. The system should optimize the use of memory and processor resources without sacrificing detection accuracy.

 

  • Technical Specifications
  1. Video Resolution: Developers need to consider the video input resolution supported by the Milk-V Duo.
  2. Memory Usage: Optimize memory allocation while ensuring algorithm efficiency, to not exceed the 256MB memory limit.

 

  • Acceptance Criteria
  1. The developed human figure detection algorithm should run stably on the Milk-V Duo, accurately identifying and tracking human figures in the video.
  2. The video stream processing function should realize smooth video capture and display, with no significant delay.
  3. Functional Testing: The system should pass at least 10 hours of video stream testing, ensuring stable operation in different scenarios without significant bugs.
  4. Performance Testing: After running continuously for 1 hour, the system should maintain stable detection accuracy, with CPU usage not exceeding 70%, and memory usage not exceeding 200MB.
  5. The submitted system should include complete source code, algorithm implementation, and necessary documentation for subsequent maintenance and optimization.