How dhiwinsamrich/yolov8_blueprint_detector works
Project type and purpose
• This is a monolithic, full-stack machine-learning inference service:
• FastAPI-based backend exposing JSON and interactive UI endpoints
• YOLOv8 object-detection model for doors and windows
• Static web frontend (drag-and-drop) for user interaction
• Docker-ready deployment with Docker Compose
File structure
• Top-level files and directories
• app.py / new_version.py: FastAPI application entrypoint
• static/, frontend.html: Web UI assets (HTML/CSS/JS)
• models/, yolov8n.pt: Trained YOLO weight files
• datasets/, train.py: Training pipeline and labeled image sets
• runs/: Inference and training run artifacts
• Dockerfile, docker-compose.yml, Start.sh: Container orchestration
• config.yaml, render.yaml: Runtime configuration parameters
• requirements.txt: Python dependencies
• Patterns and design choices
• Layered architecture: presentation layer (static/UI), application layer (FastAPI endpoints), model layer (YOLOv8 inference), persistence (model files, logs)
• Model-loading strategy: custom weights → fallback weights
• Supports two clients: HTML drag-and-drop and direct API calls (curl/Postman)
README for additional insights
• Docker instructions: single service exposing port 8000
• Model loading fallback logic
• Training evidence and labeling format
Essential elements
• Component labels (e.g., “FastAPI Server”, “YOLOv8 Inference Engine”)
• Configuration and model files as data stores