[Fall Detection GUI with YOLOv5 Python]

---

Introduction

• Class: standing, sitting, falling.
• Train set: 715 images
• Validation set: 377 images

Dependencies and Installation

• python=3.6.10
• PyTorch=1.7.1
• NVIDIA GPU=GTX1660Ti + CUDA=10.2
• numpy=1.19.5
• matplotlib=3.3.4
• pillow=8.4.0
• opencv-python=3.3.0.9

Preparation

Prepare two cameras and Trained YOLOv5 weights.

FILE STRUCTURE

Fall Detection UI with YOLOv5 (Python)
|-- README.md
|-- api
    |-- camera.py
|-- Detection results
    |-- 0.jpg
    |-- 1.jpg
    |-- Left_camera_location.txt
    |-- Right_camera_location.txt
|-- Photographed images
    |-- 0.jpg
    |-- 1.jpg
|-- YOLOv5
    |-- models
    |-- runs
        |-- train
            |-- 7_496124
                |-- weights
                    |-- best.pt
                    |-- last.pt
    |-- utils
    |-- detect.py
    
|-- front.ico
|-- UI.py

Metrics update

Methods	Precision	Recall	mAP(0.5)	mAP(0.5:0.95)
YOLOv5m	0.8039	0.8259	0.8462	0.6068

UI

Samples

自動記錄跌倒者之影像(座標)與時間

物件座標

站著

坐著

跌倒