---
license: apache-2.0
tags:
- image-quality-assessment
- image-quality
- computer-vision
- transformer
task_categories:
- image-classification
language:
- en
library_name: transformers
---

# Q-SiT: Subjective Image Quality Assessment

A deep learning model for subjective image quality assessment based on the Q-SiT architecture.

## Quick Start

### 1. Clone the repository
```bash
git clone https://huggingface.co/gongnq/SubjectiveIQE
cd SubjectiveIQE
```

### 2. Install dependencies
```bash
pip install -r requirements.txt
```

### 3. Usage

#### Web Interface
Run the web application for interactive image quality assessment:
```bash
python UI.py
```

#### Batch Testing
Test multiple images at once:
```bash
python test_images.py
```


## Model

The model files are included in this repository and will be automatically downloaded when you clone from Hugging Face.

## Dataset
### Download Full Datasets

- **Koniq/SPAQ/Q-Instruct**: https://huggingface.co/datasets/q-future/Q-Instruct-DB/blob/main/q-instruct-images.tar
- **LLaVA-150K**: https://huggingface.co/datasets/liuhaotian/LLaVA-Instruct-150K

## Requirements

- Python 3.8+
- PyTorch
- Transformers
- Other dependencies listed in `requirements.txt`