https://github.com/jjshay/art-crop-system
AI-powered artwork detection and cropping with multi-tier analysis
https://github.com/jjshay/art-crop-system
ai artwork automation computer-vision image-processing opencv pillow python
Last synced: 5 months ago
JSON representation
AI-powered artwork detection and cropping with multi-tier analysis
- Host: GitHub
- URL: https://github.com/jjshay/art-crop-system
- Owner: jjshay
- License: mit
- Created: 2026-01-08T20:01:08.000Z (6 months ago)
- Default Branch: main
- Last Pushed: 2026-01-12T02:40:08.000Z (5 months ago)
- Last Synced: 2026-01-12T04:43:14.364Z (5 months ago)
- Topics: ai, artwork, automation, computer-vision, image-processing, opencv, pillow, python
- Language: Python
- Size: 123 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 4
-
Metadata Files:
- Readme: README.md
- Changelog: CHANGELOG.md
- Contributing: CONTRIBUTING.md
- License: LICENSE
- Security: SECURITY.md
Awesome Lists containing this project
README
# Art Crop System
[](https://codecov.io/gh/jjshay/art-crop-system)
**AI-powered artwork detection and cropping - automatically extract art from photos with perfect borders.**
[](https://asciinema.org/a/Z0Qp6keIpvrnztvS)
[](https://asciinema.org/a/Z0Qp6keIpvrnztvS)
---
## What Does This Do?
When you photograph artwork, you get the frame, wall, shadows, etc. This tool:
1. **Detects the artwork** in your photo using AI vision
2. **Finds the exact edges** of the art piece
3. **Crops perfectly** - no extra background
4. **Generates detail shots** - corners, signature, texture
5. **Validates quality** - checks for blur, lighting, angle issues
**Before:** Photo of framed art on a wall
**After:** Clean, perfectly cropped artwork image + 8 detail shots
---
## Quick Start
```bash
# Clone the repo
git clone https://github.com/jjshay/art-crop-system.git
cd art-crop-system
# Install dependencies
pip install -r requirements.txt
# Run the interactive demo
python demo.py
# Or run the visual showcase
python showcase.py
# Crop sample wall photo
python ai_art_crop_system.py examples/wall_photo.jpg --output output/
```
### Sample Files
- `examples/wall_photo.jpg` - Photo of artwork on wall (ready to crop!)
- `examples/crop_config.json` - Configuration options
- `sample_output/crop_report.json` - Example detection results
---
## Architecture
```mermaid
flowchart TB
subgraph Input
A[Wall Photo] --> B[Image Loader]
end
subgraph Tier1["Tier 1: Fast Processing"]
B --> C[rembg Background Removal]
C --> D{Clean Result?}
end
subgraph Tier2["Tier 2: Multi-AI Analysis"]
D -->|No| E1[GPT-4 Vision]
D -->|No| E2[Claude Vision]
D -->|No| E3[Gemini Vision]
D -->|No| E4[Grok Vision]
E1 & E2 & E3 & E4 --> F[Boundary Consensus]
end
subgraph Tier3["Tier 3: CV Fallback"]
F --> G{Consensus?}
G -->|No| H[OpenCV Edge Detection]
end
subgraph Output
D -->|Yes| I[Crop Engine]
G -->|Yes| I
H --> I
I --> J[Detail Shot Generator]
J --> K[Quality Validator]
K --> L[Final Outputs]
end
style A fill:#e1f5fe
style L fill:#c8e6c9
style C fill:#fff9c4
style F fill:#f3e5f5
```
## How It Works
```
Original Photo (art on wall)
│
▼
┌─────────────────────────────────────────┐
│ TIER 1: rembg (Fast) │
│ Background removal for initial crop │
└─────────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────┐
│ TIER 2: Multi-AI Vision Analysis │
│ GPT-4V + Claude + Gemini + Grok │
│ Consensus on exact artwork boundaries │
└─────────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────┐
│ TIER 3: CV Edge Detection │
│ OpenCV fallback for edge cases │
└─────────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────┐
│ TIER 4: Quality Check │
│ Blur detection, lighting, angle │
└─────────────────────────────────────────┘
│
▼
Perfect Crop + Detail Shots
```
---
## Output: 8 Detail Shots
For each artwork, the system generates:
| Shot | Shows |
|------|-------|
| Full | Complete artwork |
| Top-Left | Upper left corner detail |
| Top-Right | Upper right corner detail |
| Bottom-Left | Lower left corner + signature area |
| Bottom-Right | Lower right corner |
| Center | Central detail/texture |
| Signature | Detected signature closeup |
| Edition | Edition number if present |
---
## Features
- **Multi-AI Consensus**: Uses 4 AI models to agree on crop boundaries
- **Signature Detection**: Automatically finds and highlights artist signatures
- **Edition Number Detection**: Locates "23/100" style edition markings
- **Quality Scoring**: Rates image quality (lighting, focus, angle)
- **Batch Processing**: Process entire folders automatically
- **Google Drive Integration**: Upload results directly to Drive
---
## Setup
### Basic (No AI)
```bash
pip install -r requirements.txt
python demo.py # Uses rembg only
```
### Full AI Mode
```bash
# Copy and configure .env
cp .env.example .env
nano .env
# Add your API keys
OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...
GOOGLE_API_KEY=AIza...
```
---
## Files
| File | Purpose |
|------|---------|
| `ai_art_crop_system.py` | Main multi-tier cropping system |
| `gdrive_gallery_quality_crops.py` | Google Drive integration |
| `crop_quality_validator.py` | Image quality assessment |
| `art_crop_pipeline_final.py` | Production pipeline |
| `demo.py` | Demo without API keys |
---
## License
MIT - Use freely for your artwork processing!