How samamostafa03/art-critique-and-creation-tool works
Project type and purpose
• From the README and file tree, Art-Vision is an AI-powered microservices system exposing two sets of capabilities via REST APIs:
• Art Creation (cartoon, style-transfer, watercolor)
• Art Critique (description generation, genre/style classification, scoring)
• Each model lives in its own folder with a Dockerfile, requirements.txt, a FastAPI-style app.py, and notebooks.
File structure
• Top-level:
• AI-Models/Creation-Models/…
• AI-Models/Critique-Models/…
• README.md, Art-Vision-documentation.pdf
• Under each model directory:
• Dockerfile + .dockerignore → containerized service
• app.py → FastAPI (or similar) entrypoint exposing endpoints
• requirements.txt → Python dependencies
• No monolithic backend folder: each model is its own service
Insights from the README
• Uses FastAPI (invoked via uvicorn … --reload)
• Exposes a Postman/Swagger spec
• Relies on pre-trained networks (e.g. DenseNet-121) and in-folder model artifacts or JSONs
Main components
a. Client Layer
• Web or CLI client (not in repo)
• Postman / Third-party integrators
b. API Gateway / Load Balancer
• Single entrypoint (FastAPI routing layer or API Gateway)
• Routes incoming requests to the appropriate model service
c. Model Microservices (six services)
• Art-Creation: Cartoon Service, Style-Transfer Service, Watercolor Service
• Art-Critique: Description-Generator Service, Genre-Style Classifier Service, Scoring Service
d. Model Storage
• Pre-trained weights & JSON artifacts (mounted as volumes in each container)
e. Logging / Metrics (optional)
• Centralized logging/monitoring (e.g., ELK, Prometheus)
f. Documentation
• Swagger / OpenAPI UI (served by API Gateway)
• PDF user guide
Relationships and data flow
• Client → API Gateway (HTTP call)
• API Gateway → forwards to specific Dockerized Model Service
• Model Service → loads model weights → performs inference → returns JSON/image to API Gateway → API Gateway forwards to Client
• All services push logs/metrics to centralized Logging/Monitoring
Architectural patterns and principles
• Microservices: each AI model isolated in its own container
• Twelve-factor app: separate code, config, dependencies via requirements.txt and Dockerfiles
• API gateway pattern: central routing and security layer
• Containerization: all inference code runs in Docker, enabling consistent environments
Technologies and frameworks
• Python 3.x, FastAPI (uvicorn server)
• Deep learning backbones (DenseNet-121, custom CNNs)
• Docker & Docker Compose / Kubernetes for orchestration
• Jupyter notebooks for model experimentation