---
title: CIET
emoji: 👀
colorFrom: yellow
colorTo: purple
sdk: gradio
sdk_version: 5.18.0
app_file: app.py
pinned: false
license: apache-2.0
short_description: Comprehensive Image Evaluation Tool
---

# Comprehensive Image Evaluation Tool

This tool combines multiple image evaluation models into a single application with a user-friendly interface for analyzing and reviewing images.

## Features

- **Batch Processing**: Upload multiple images at once for efficient evaluation
- **Multiple Models**: Combines evaluations from several aesthetic prediction models:
  - ShadowLilac's aesthetic-shadow-v2
  - WaifuScorer V4
  - CafeAI's aesthetic, style and waifu classifiers
  - Anime Aesthetic predictor
- **Comprehensive Analysis**: Get detailed metrics for each image
- **Results Table**: View results sorted by score with image previews
- **Export**: Save results to CSV for further analysis
- **Single Image Mode**: Evaluate individual images and get detailed results

## Installation

1. Clone this repository:
   ```
   git clone [repository-url]
   cd image-evaluation-tool
   ```

2. Install required dependencies:
   ```
   pip install -r requirements.txt
   ```

3. Run the application:
   ```
   python app.py
   ```

## Usage

### Batch Processing

1. Launch the application
2. Use the file upload panel to select multiple images
3. Adjust the HQ threshold if needed (default 0.5)
4. Click "Process Images"
5. View results in the table sorted by average score
6. Click "Export Results to CSV" to save the data

### Single Image Evaluation

1. Scroll down to the Single Image Evaluation section
2. Upload an image
3. Click "Evaluate"
4. View detailed metrics and style information

## Models Information

- **ShadowLilac** (0-1): General aesthetic quality assessment
- **WaifuScorer** (0-10): Specialized for anime-style images
- **CafeAI** (0-1): Style classification and aesthetic assessment
- **Anime Aesthetic** (0-10): Specialized for anime/manga art

## Output Folders

- `output/hq_folder`: Images that meet or exceed the threshold
- `output/lq_folder`: Images that score below the threshold

## Requirements

- Python 3.8+
- CUDA-compatible GPU recommended for faster processing
- ~4GB of disk space for model downloads (first run)