# Sanity: Synthetic Image Generation for Computer Vision

> The "Vending Machine" for Computer Vision engineers. Trade money for headache removal.

## What is Sanity?
Sanity is a specialized tool that generates, auto-labels, and formats synthetic image datasets for computer vision applications. Instead of scraping images and manually drawing bounding boxes, you describe what you want and get a ready-to-train dataset.

## How it works
1. **Setup**: Define your prompt (e.g., "Drone in snow") and select your format (YOLO, COCO, or Pascal VOC).
2. **Payment**: Simple pay-per-image model ($0.10/image).
3. **Batch Processing**: Jobs run asynchronously using AI for image generation and auto-labeling.
4. **Auto-Refund**: Failed images and labels are automatically refunded.
5. **Export**: Download a perfectly formatted .zip file containing images and labels ready for training.
6. **Verify**: Run the included Python visualizer script to verify bounding boxes.

## Key Features
- **AI Auto-Labeling**: Extremely accurate bounding box generation.
- **Formats**: YOLO, COCO, and Pascal VOC.
- **Prepaid Wallet**: No subscriptions. Minimum deposit $5.
- **10-5000 images**: Scalable batch sizes for any project.
- **Visualizer Script**: Python tool included in every download to verify labels.

## Pricing
- $0.10 per image (generation + labeling + formatting)
- Failed images/labels automatically refunded
- Minimum deposit: $5

## Use Cases
- Object detection model training
- Edge case data augmentation
- Rapid prototyping of CV models
- Privacy-compliant synthetic data

## Formats Supported
- **YOLO**: .txt files with normalized coordinates
- **COCO**: JSON annotation format
- **Pascal VOC**: XML annotation format

## Contact
Website: https://sanitycv.com