Mar 30
Loading video player...
By Yuwan Nanda Adyatma | Class 109 In this video, I walk through an end-to-end computer vision pipeline I built from scratch using Docker Compose — running 5 microservices that work together to detect objects in dashcam footage and output real-time driving decisions. ───────────────────────────────────── 🔧 WHAT'S COVERED ───────────────────────────────────── - Docker Compose architecture — 5 containerized microservices - Service communication over internal Docker bridge network - Image preprocessing pipeline (resize, contrast, sharpen) - YOLOv8s object detection with confidence filtering - Spatial zone filtering & risk-based decision logic - Real-time dashboard built with vanilla HTML/CSS/JS ───────────────────────────────────── 🛠️ TECH STACK ───────────────────────────────────── - Python + FastAPI + Uvicorn - Ultralytics YOLOv8s - Docker + Docker Compose - Pillow + NumPy (image processing) - HTML5 Canvas API (client-side video frame extraction) - Vanilla JavaScript (no framework) ───────────────────────────────────── 📁 PROJECT STRUCTURE ───────────────────────────────────── sensor/ → frame buffer service preprocessing/ → image standardization yolo_detection/ → AI inference engine decision/ → risk scoring & action logic frontend/ → dashboard UI ───────────────────────────────────── #docker #computervision #yolov8 #python #fastapi #machinelearning #objectdetection #dockercompose #softwaredevelopment #mlops