65 lines
1.8 KiB
Markdown
65 lines
1.8 KiB
Markdown
# AI Design & AI Fine-tuning
|
|
|
|
This repository contains two taks:
|
|
|
|
1. AI Design Analyzer - Convert wireframes/screenshots to detailed UI requirements and React code
|
|
2. BaaS(Backend As A Service) AI Fine-tuning - Fine-tune and generate backend endpoints based on existing schemas
|
|
|
|
## Project Overview
|
|
|
|
### AI Design Analyzer
|
|
|
|
- Takes wireframe/website screenshots as input
|
|
- Analyzes and extracts UI requirements including:
|
|
- Typography
|
|
- Color schemes
|
|
- Button & Input styles
|
|
- Icons & Images
|
|
- Layout & Sections
|
|
- Interactive elements
|
|
- Generates production-ready React components
|
|
|
|
### AI BaaS Endpoint Fine-tuner
|
|
|
|
- Fine-tunes existing backend endpoints
|
|
- Generates new endpoints based on schema definitions
|
|
- Reuses existing services when applicable
|
|
- Creates new services only when necessary
|
|
|
|
## 🚀 Getting Started
|
|
|
|
### Prerequisites
|
|
|
|
- Python 3.10+
|
|
|
|
## 📋 Model Training
|
|
|
|
### AI Design Model
|
|
|
|
The AI Design analyzer uses a model fine-tuned on UI wireframes and designs. The model is trained to detect:
|
|
|
|
- UI elements and their relationships
|
|
- Style properties and patterns
|
|
- Layout structures
|
|
- Interactive components
|
|
|
|
**Note**: Take any samples of data(Wireframe Image) and then do the above. For example: You can take example of UI from there: [https://dribbble.com/tags/ui](https://dribbble.com/tags/ui)
|
|
|
|
### Basically the flow will look like that:
|
|
|
|
|
|
|
|
### AI BaaS Fine-tuning Model
|
|
|
|
The BaaS fine-tuning uses a CodeGen-based model trained to:
|
|
|
|
- Understand existing endpoint patterns
|
|
- Recognize schema relationships
|
|
- Generate consistent API endpoints
|
|
- Maintain service architecture
|
|
**Note**: For fine-tuning use the data in the `/ai_finetunining/dataset` directory
|
|
|
|
### The flow looks like this:
|
|
|
|
|