Basic Capstone Design : BELIEVE

---

Vision Based Edge Intelligence System for Safety Management on Industrial Environment

Team AASC(Anywhere Anytime Security with CCTV) , Mar. 2024 ~ Jun. 2024

---

Team members

Name	Role	Email
Seungeun Kang	Team Leader, AI engineer	haun620@kyonggi.ac.kr
Kibum Park	Hardware Engineer	rlqja530@kyonggi.ac.kr
Gunwoo Park	Frontend programmer	dbel56@kyonggi.ac.kr

Software Stacks

---

Project Motivation

1. Rasing awareness of safety management

• Serious Accidents Punishment Act, Korean Law ( from 2021 )

- Accidents at industrial sites, construction sites, etc. cause significant loss of life and property damage.

2. Problems with existing AI-based safety management systems

• AI inference is executed on “Server”
• Requires high traffic and high performance GPU servers
• We focused on “edge intelligence” technologies

	Warn notification	Portable	Using Sensor data	Real time inference	Edge inference	Physical Volume
Existing tech	yes	yes	yes	yes	no	big
BELIEVE	yes	yes	yes	yes	yes	small

---

Implementation

1. Overall System

2. Object Detection Model (Safety Helmet Detection)

• Dataset : Kaggle - Safety Helmet Detection Dataset (link)
• Model : Yolo v7 Tiny (link)
• Detection From Test dataset

---

Outputs

1. Conference Paper

• Publication conference paper in Korean Institute of Information Technology Summer Conference (May. 2024)
• 🏆 Best Paper Award(Gold prize) in Korean Institute of Information Technology Summer Conference (May. 2024)

2. Poster

• Poster presentation in Basic Capstone Design competition, Kyonggi University (Jun. 2024)
• 🏆 Excellence Award(Gold prize) in Basic Capstone Design competition, Kyonggi University (Jun. 2024)