This document summarizes a series of YOLOv8 experiments and implementations I’ve been conducting to build toward a robust real-time object detection and recognition system, which will soon be integrated with the RealSense D435i camera.

Each project explores a different aspect of detection: from model experimentation, webcam integration, to applied use cases like people and PPE detection with real-time inference.

1. YOLOv8 Object Detection Experiment

GitHub Repository: YOLOv8 Detection Experiment

Description:

This project explores various YOLOv8 model sizes (n, s, m, l, x) on different datasets for object detection tasks. It focuses on fine-tuning YOLOv8 for speed vs accuracy trade-offs using pretrained models and local test sets.

Features:

• Comparison of model inference speeds
• Training on custom datasets
• Visualization of bounding boxes

Visual

2. YOLOv8 Object Detection with Webcam

📂 GitHub Repository: YOLOv8 with Webcam

Description:

This implementation connects a webcam (USB camera) to YOLOv8 for real-time object detection. It builds on previous models and adds user interface enhancements for frame display.

Features:

• Real-time detection via OpenCV + YOLOv8
• Adjustable confidence thresholds