Art-Vision: Art Critique and Creation Tool

Art-Vision is an AI-powered tool designed to offer two core functionalities:

• Art Creation - Generates original artwork based on user-defined description, styles, or parameters, providing creative inspiration art pieces.
• Art Critique - Analyzes an artist's work and provides a critique of various elements, such as color palettes, composition, texture, and style, which can help artists improve or refine their works.

For a detailed explanation of the project, please refer to the full documentation: 📌 Project Documentation (PDF)

AI Models

1. Multi-Class Classification Model

• Backbone Model: Convnext (fine-tuned with transfer learning)
• Dataset: WikiArt
• Task: Predicts both art style and genre from an artwork image

2. Regression Model

• Backbone Model: CLIP+MLP (fine-tuned with transfer learning)
• Dataset: APDDv2
• Task: Predicts color harmony, Texture, composition and aesthitic scores from an artwork image

3. Image-to-Text Generation Model

• Backbone Model: BLIP-Base (fine-tuned with transfer learning)
• Dataset: WikiArt
• Task: Generates descriptive caption of an artwork image

4. Text-to-Image Generation Model

• Backbone Model: Stable Diffusion-XL (fine-tuned with Lora Enabled)
• Dataset: Custom dataset
• Task: Generates images in watercolor style based on text prompt

📄 Postman API Documentation

👉 Fast API Documentation

Running the FastAPI Backend

uvicorn app:app --reload